The Taub Faculty of Computer Science Events and Talks

Secure execution technologies, such as Intel SGX, offer an attractive solution for protecting one's private data in the public cloud environment. In this talk, we will explore how SGX mitigates various attack surfaces and what are the caveats of naively using the technology to protect applications. Specifically, we will discuss the performance implications of SGX on common applications and understand what are the new bottlenecks created by SGX, which may lead to a 5x performance d...
[ Full version ]

A novel yacht hull design framework and a generative design technique for CAD products will be explained in this talk. A new design framework for the parametric design and shape modification of a yacht hull will be outlined first. In this framework, the hull is divided into three regions (entrance, middle and run) and each region is represented separately using Coons patches. Shape operators helps designers to modify the given hull shape while considering some quality criteria suc...
[ Full version ]

Many software systems rely on data-driven models to make decisions. Examples include self-driving cars, malware detection and aircraft collision avoidance detection. Unfortunately, data-driven models often do not generalize well on unseen examples, despite showing high accuracy on test sets. This was demonstrated by showing how to fool these models using adversarial examples. Such adversarial examples may result in disastrous consequences in safety-critical systems that rely on th...
[ Full version ]

Multiple surveys show that SQL and relational databases remain the most common tools used by data scientists. But can we fully trust them? We give a few examples showing unexpected and counterintuitive behavior of even simple SQL queries that make one question analytics results obtained from relational DBMSs. The talk will then give a quick overview of two lines of work that attempt to overcome these problems. One concerns a formal semantics of SQL, to at least eliminate the eleme...
[ Full version ]

Let G=(V,E) be an n-vertices m-edges directed graph. Let s∈ V be any designated source vertex. We address the problem of single source reachability (SSR) from s in presence of failures of vertices/edges. We show that for every k≥ 1, there is a subgraph H of G with at most 2^k n edges that preserves the reachability from s even after the failure of any k edges. Formally, given a set F of k edges, a vertex u∈ V is reachable from s in G∖ F if a...
[ Full version ]

Callbacks are essential in many programming environments, but drastically complicate program understanding and reasoning because they allow to mutate object’s local states by external objects in unexpected fashions, thus breaking modularity. The famous DAO bug in the cryptocurrency framework Ethereum, employed callbacks to steal $150M. We define the notion of Effectively Callback Free (ECF) objects in order to allow callbacks without preventing modular reasoning. ...
[ Full version ]

In this talk we will explore deep neural networks from a dynamical systems point of view. We will show that the learning problem can be cast as a path planning problem with PDE constraint. This opens the door to conventional computational techniques that can speed up the learning process and avoid some of the local minima. ...
[ Full version ]

Open Access (OA) refers to various forms of research outputs that are openly available online and are free of many restrictions on access and usage. Open Access publishing has become more and more prevalent in the last decade. The benefits to the academic community and the general public are clear: scholarly publications can be easily accessed through popular free search engines, like Google Scholar. Knowledge is spread more widely and contributes to scientific progress. ...
[ Full version ]

Machine learning has recently been revolutionized by the introduction of Deep Neural Networks. However, from a theoretical viewpoint these methods are still poorly understood. Indeed the key challenge in Machine Learning today is to derive rigorous results for optimization and generalization in deep learning. In this talk I will present several tractable approaches to training neural networks. At the second part I will discuss a new sequential algorithm for decision making ...
[ Full version ]

We consider the following problem introduced by [Censor-Hillel et al., DISC 2016]. Design a distributed algorithm (called an $\epsilon$-tester) that tests whether the network over which the algorithm is running satisfies a given property (e.g., acyclic, bipartite) or is $\epsilon$-far from satisfying the property. If the network satisfies the property, then all processors must accept. If the network is $\epsilon$-far from satisfying the property, then (with probabi...
[ Full version ]

Over the past two decades, machine learning has rapidly evolved and emerged as a highly influential discipline of computer science and engineering. One of the pillars of machine learning is mathematical optimization, and the connection between the two fields has been a primary focus of research. In this talk, I will present two recent works that contribute to this study, focusing on online learning---a central model in machine learning for sequential decision making and lear...
[ Full version ]

The minimum dominating set problem in a graph $G = (V, E)$ asks to find a minimum size set $D \subseteq V$ such that every vertex not in $D$ is adjacent to at least one member of $D$. It is a well known NP-hard optimization problem that was studied from the 1950s onwards. A greedy algorithm for the minimum dominating set problem yields a $O(\log \Delta)$-approximation ratio, where $\Delta$ is the maximum degree in $G$, and this is the best possible, assuming $P \neq...
[ Full version ]

Local-to global point cloud registration is a challenging task due to the substantial differences between these two types of data, and the different techniques used to acquire them. Global clouds cover large-scale environments and are usually acquired aerially, e.g., 3D modeling of a city using Airborne Laser Scanning (ALS). In contrast, local clouds are often acquired from ground level at a much smaller range, using Terrestrial Laser Scanning (TLS). The differences are typically ...
[ Full version ]

The Border Gateway Protocol (BGP) is a crucial part of the Internet infrastructure. However, it was developed in the 1980s with limited concern for security. In particular, its lack of authentication makes it vulnerable to the so-called prefix hijacking attack. In this attack, a malicious or compromised BGP router announces a route to an IP prefix it does not own. Consequently, packets destined to this prefix are actually forwarded to the attacker. A special case of this attac...
[ Full version ]

See attached document....
[ Full version ]

Digital geometry processing (DGP) is one of the core topics of computer graphics, and has been an active line of research for over two decades. On one hand, the field introduces theoretical studies in topics such as vector-field design, preservative maps and deformation theory. On the other hand, the tools and algorithms developed by this community are applicable in fields ranging from computer-aided design, to multimedia, to computational biology and medical imaging. ...
[ Full version ]

CS invites all-degree students to a Research Presentation Event including lectures in various topics by faculty members: Miri Ben-Chen - Computing with 3D Surfaces Eli Ben-Sasson - Zero Knowledge Proofs for Cryptocurrencies Alex Bronstein - Computer Vision, Machine Learning, and Intelligent Systems Shachar Itzhaky - Program Synthesis and Derivation by Refin...
[ Full version ]

Ramsey theory assures us that in any graph there is a clique or independent set of a certain size, roughly logarithmic in the graph size. But how difficult is it to find the clique or independent set? If the graph is given explicitly, then it is possible to do so while examining a linear number of edges. If the graph is given by a black-box, where to figure out whether a certain edge exists the box should be queried, then a large number of queries must be issued. But what if one...
[ Full version ]

This seminar presents a fault-tolerant distributed operating system designed to harness the massive parallelism in many-core (1,000-10,000+) distributed shared memory processors. In order to scale efficiently and reliably as cores count rapidly increase while their reliability decrease, the new operating system provides fault-tolerant task-level parallelism using coarse-grained data-flow principles. Combining message passing and shared memory, a wait-free decentralized execution e...
[ Full version ]

It is widely recognized that a major disruption is under way in computer hardware as processors strive to extend, and go beyond, the end-game of Moore's Law. This disruption will further extend current software trends towards increasing scales and ubiquity of parallelism, to include new levels of "extreme heterogeneity" motivated by heterogenous processor and memory hierarchies, near-memory computation structures, and even Non-von Neumann computing elements.. Since all software ...
[ Full version ]

A combination of supply-voltage scaling, GHz+ operating frequencies and multi-core execution makes power-delivery a critical challenge for high-end processing systems. Previous work on Power-Delivery Network monitoring approaches consume expensive pad resources or suffer from design-time and run-time overheads. In this talk, we present a non-intrusive power delivery network monitoring methodology based on sensing modulations in the emanated CPU electromagnetic radiation due to dyn...
[ Full version ]

I will talk about two recent results concerning problems motivated by human cognition. (i) In the first part I will talk about modeling of phenomena in perceptual learning. Building on the powerful tools currently available for the training of Convolution Neural Networks (CNN), networks whose original architecture was inspired by the visual system, we revisited some of the open computational questions in perceptual learning. We first replicated two representative sets of perceptua...
[ Full version ]

We are happy to invite your to the second of series of meetings on career and job seeking which will be held at CS. Orly Militzer, HR Speaker at Dell EMC ScaleIO, will give a talk on How to start looking for a job: - How to look and find a proper job - Improving your chances to find the first job - Principles and examples to writing compatible CV - Preps for job interviews - Principles of effective self presentation in job interviews - D...
[ Full version ]

Deep Neural Networks are succeeding at a range of natural language tasks such as machine translation and text summarization. Recently, the interdisciplinary field of "big code" promises a new set of learnable statistical static analyses. While machine learning tasks on source code have been considered recently, most work in this area does not attempt to capitalize on the unique opportunities offered by its known syntax and structure. In this talk, I discuss how graph neural networ...
[ Full version ]

We give the first construction of high-rate locally list-recoverable codes. List-recovery has been an extremely useful building block in coding theory, and our motivation is to use these codes as such a building block. In particular, our construction gives the first capacity-achieving locally list-decodable codes; the first capacity achieving globally list-decodable codes with nearly linear time list decoding algorithm; and a randomized construction of binary codes on the Gilbert...
[ Full version ]

Monitoring tasks, such as anomaly and DDoS detection, require identifying frequent flow aggregates based on common IP prefixes. These are known as hierarchical heavy hitters (HHH), where the hierarchy is determined based on the type of prefixes of interest in a given application. The per-packet complexity of existing HHH algorithms is proportional to the size of the hierarchy, imposing significant overheads. In this talk, I will present a constant update time algorithm ...
[ Full version ]

Network Function Virtualization (NFV) is an emerging networking paradigm where network functions are located at Commercially-Off-The-Shelf (COTS) servers distributed across a network. This paradigm is a major turning point in the evolution of networking, as it introduces high expectations for both agile and much more economical network services. However, in order to achieve high utilization and low cost, one must deploy resource allocation algorithms that are aware of the dynamici...
[ Full version ]

We give the first construction of high-rate locally list-recoverable codes. List-recovery has been an extremely useful building block in coding theory, and our motivation is to use these codes as such a building block. In particular, our construction gives the first capacity-achieving locally list-decodable codes; the first capacity achieving globally list-decodable codes with nearly linear time list decoding algorithm; and a randomized construction of binary codes on the Gilbert...
[ Full version ]

Models of uncertain preferences, such as Mallows, have been extensively studied due to their plethora of application domains. In a recent work, a conceptual and theoretical framework has been proposed for supporting uncertain preferences as first-class citizens in a relational database. The resulting database is probabilistic, and, consequently, query evaluation entails inference of marginal probabilities of query answers. In this paper, we embark on the challenge of a prac...
[ Full version ]

Over the next few decades, we are going to transition to a new economy where highly complex, customizable products are manufactured on demand by flexible robotic systems. This change is already underway in a number of fields. 3D printers are revolutionizing production of metal parts in aerospace, automotive, and medical industries. Whole garment knitting machines allow automated production of complex apparel and shoes. Manufacturing electronics on flexible substrates opens the doo...
[ Full version ]

The recurring context in which objects appear holds valuable information that can be employed to predict their existence. This intuitive observation indeed led many researchers to endow appearance-based detection results with explicit reasoning about context. The underlying thesis suggests that with stronger contextual relations, the better improvement in detection capacity one can expect from such a combined approach. In practice, however, the observed improvement in many case is...
[ Full version ]

Cryo-EM is an imaging technology that is revolutionizing structural biology; the Nobel Prize in Chemistry 2017 was recently awarded to Jacques Dubochet, Joachim Frank and Richard Henderson "for developing cryo-electron microscopy for the high-resolution structure determination of biomolecules in solution". Cryo-electron microscopes produce a large number of very noisy two-dimensional projection images of individual frozen molecules. Unlike related methods, such as computed tomog...
[ Full version ]

The influence of the k'th coordinate on a Boolean function f:{0,1}^n -> {0,1} is the probability that flipping x_k changes the value f(x). The total influence I(f) is the sum of influences of the coordinates. The classical `Junta Theorem' of Friedgut (1998) asserts that if I(f) ...
[ Full version ]

Object classification and more generally - semantic perception, are an important aspect in situational awareness in autonomous systems. Recent advances in visual information processing have enabled the use of rich semantic information in critical systems, spurring demand for robust, uncertainty-aware semantic perception. The integration of semantic information with noisy spatial (pose, world geometry) information results in mixed - continuous and discrete state belief, ofte...
[ Full version ]

I will present a novel comprehensive theory of large scale learning with Deep Neural Networks, based on the correspondence between Deep Learning and the Information Bottlneck framework. The theory is based on the following components: (1) rethinking Learning theory. I will prove a new generalization bound, the input-compression bound, which shows that compression of the input variable is far more important for generalization than the dimension of the hypothesis class,...
[ Full version ]

Co-occurrence Filter (CoF) is a boundary preserving filter. It is based on the Bilateral Filter (BF) but instead of using a Gaussian on the range values to preserve edges it relies on a co-occurrence matrix. Pixel values that co-occur frequently in the image (i.e., inside textured regions) will have a high weight in the co-occurrence matrix. This, in turn, means that such pixel pairs will be averaged and hence smoothed, regardless of their intensity differences. On the other...
[ Full version ]

Microsoft will hold an event of lectures, drinks and snacks on Monday evening, November 20, 2017, between 17: 00-18: 30, in the lobby of the Taub Building. The event will include meetings with senior officials, researchers and engineers who will offer you employment opportunities and will guide you how to achieve them, as well as a lecture by the director general of the R & D center, Mr. Yoram Yaakobi ...
[ Full version ]

The goal of this work is to develop collaborative mechanisms which help people gain understanding of complex phenomena. We start by presenting an online, collaborative system for the study of child development. Moving from practice to theory, we proceed by introducing an abstract mathematical model for user retention, facilitating the design of efficient crowd collaboration systems. The first part of the work is dedicated to the Baby CROINC (CROwd INtelligence Curation) system, w...
[ Full version ]

Random matrices have become a very active area of research in the recent years and have found enormous applications in modern mathematics, physics, engineering, biological modeling, and other fields. In this work, we focus on symmetric sign (+/-1) matrices (SSMs) that were originally utilized by Wigner to model the nuclei of heavy atoms in mid-50s. Assuming the entries of the upper triangular part to be independent +/-1 with equal probabilities, Wigner showed in his pioneering wor...
[ Full version ]

We introduce a novel lower bound technique for distributed graph algorithms under bandwidth limitations. We define the notion of fooling views and exemplify its strength by proving two new lower bounds for triangle membership in the CONGEST(B) model: (i) Any 11-round algorithm requires B≥cΔlogn for a constant c>0c. (ii) If B=1, even in constant-degree graphs any algorithm must take Ω(log∗n) rounds. The implication of the former is the first proven separation between the LOCA...
[ Full version ]

We present a general approach to video understanding, inspired by semantic transfer techniques that have been successfully used for 2D image analysis. Our method considers a video to be a 1D sequence of clips, each one associated with its own semantics. The nature of these semantics -- natural language captions or other labels -- depends on the task at hand. A test video is processed by forming correspondences between its clips and the clips of reference videos with known semantic...
[ Full version ]

There is a large literature explaining why AdaBoost is a successful classifier. The literature on AdaBoost focuses on classifier margins and boosting's interpretation as the optimization of an exponential likelihood function. These existing explanations, however, have been pointed out to be incomplete. A random forest is another popular ensemble method for which there is substantially less explanation in the literature. We introduce a novel perspective on AdaBoost and random...
[ Full version ]

Gurus are individuals who claim to possess mental powers of insight and prediction that far surpass those of the average person; they compete over followers, offering them insight in return for continued devotion. Followers wish to harness a (true) Guru’s predictive power but (i) have limited attention span and (ii) doubt the Guru’s predictive advantage over them. This dynamic raises the question of follower retention: how do Gurus retain the faith of their flock in the face o...
[ Full version ]

Sparse approximation and dictionary learning have been applied with great success to several image processing tasks, often leading to state-of-the-art results. Yet, these methods have traditionally been restricted to small dimensions due to the computational constraints that these problems entail. This paradigm results in a series of inconsistencies, however, with both practical and theoretical implications. I will first review a series of algorithmic solutions to this local-globa...
[ Full version ]

In the near future, there will likely be special-purpose quantum computers with 50 or so high-quality qubits. In this talk, I'll discuss general theoretical foundations for how to use such devices to demonstrate "quantum supremacy": that is, a clear quantum speedup for *some* task, motivated by the goal of overturning the Extended Church-Turing Thesis (which says that all physical systems can be efficiently simulated by classical computers) as confidently as possible. Based on r...
[ Full version ]

When video is recorded in a studio, sound is clear of external noises and unrelated sounds. However, most video is not shot at studios. Video conferences from home or office are often disturbed by other people, ringing phones, or barking dogs. TV reporting from city streets is commonly mixed with traffic noise or sound of winds. We can exploit the visual information of face and mouth movements as seen in the video to enhance the voice of a speaker,...
[ Full version ]

The coherent nature of ultrasound imaging along with the bandlimitedness of its beamforming mechanism impose limitations on the resolution and contrast of ultrasound scans. While rarely corrigible via refinement of hardware design, the above limitations can be effectively mitigated by means of post-processing. In this case, the problem of image restoration is usually cast in the form of an inverse problem, which is subsequently solved using numerical optimization. One of the most ...
[ Full version ]

We are happy to invite your to the first of series of meetings on career and job seeking which will be held at CS. How to turn a degree to a career? Dr. YonathanYaniv, a CS graduate and algorithms and learning system researcher at YOTPO, will give a talk on: - Picking and selecting the right courses and project throughout studies. - Choosing the right track - Sorts of Student jobs The event will be held on Sunday, November 5th, 17:00, ...
[ Full version ]

For over 4 decades, the gap between computer processing speed and memory access has grown at about 50% per year, to more than 1,000x today. This provides an excellent opportunity to enhance the single-core system performance. An innovative 3D integration technology combined with re-architecting the integrated memory device is proposed to bridge the gap and enable a 1,000 x improvement in computer systems. The proposed technology utilizes processes that are widely available and cou...
[ Full version ]

In the past decade we are experiencing a massive come-back of Neural Networks that flourished in the eighties, this time with a much greater success. This return could be attributed in part to the emergence of new training techniques that allows training deep networks efficiently. Obviously, progresses in hardware have also played an important role in this come-back: Powerful graphical processing units (GPU) are very useful when matrices and vectors operations are needed, leading ...
[ Full version ]

Google team will visit CS to give a tech talk focusing on Machine Learning and to present to you various opportunities to discuss with their experts. Please pre-register. You are all invited!...
[ Full version ]

In this work, we continue the study of a new class of codes, called codes over graphs Here we consider storage systems where the information is stored on the edges of a complete directed graph with n nodes. The failure model we consider is of node failures which are erasures of all edges, both incoming and outgoing, connected to the failed node. It is said that a code over graphs is a \rho-node-erasure-correcting code if it can correct the failure of any \rho nodes in the graphs o...
[ Full version ]

The choice of hyper parameters, such as learning rate, when training Deep Neural Networks is more of art than science. However, correctly setting them is often crucial to the success of the training process. Therefore, the common practice is to try many options using various rules of thumb. As a result, the quest after the best hyper parameters is the most time consuming phase in designing the network, and often the main source of frustration. We propose two techniques to dy...
[ Full version ]

Optimistic methods for solving Reinforcement Learning problems are very popular in the literature. In practice, however, these methods show inferior performance compared to other methods, such as Posterior Sampling. We propose a novel concept of Restricted Optimism to balance the well known exploration vs. exploitation trade-off for finite-horizon MDPs. We harness Posterior Sampling to construct two algorithms in the spirit of our Restricted Optimism principle. We provide theoreti...
[ Full version ]

Computing the sum of elements over a sliding window is a textbook interview question. By storing the last window and its sum in memory, we can process elements and answer queries in constant time and near-optimal space. In this talk, I will discuss a variant of this problem where the user specifies the window size i≤n query time, and only an upper bound n is known in advance. As window sizes in practice may be large, standard practice is to settle on appro...
[ Full version ]

Remote computing services (e.g., virtualization and cloud services) offer advantages to organizations and individuals, putting at their disposal enormous computing resources while permitting them to pay only for the resources actually used. Unfortunately, such environments are prone to attacks by hackers, adversarial users of the systems, or even the owner of the service. Such attacks may address the operating system, hypervisor, VMM, or even the hardware itself. It would therefor...
[ Full version ]

CS invites all degrees students to attend the "Exposure Night" - An information fair that will present the variety of CS projects, seminars and advanced courses to be given in the 2017-2018 academic year. The event will take place on Tuesday's evening, 18:00, in the lobby of CS Taub Building. You are all invited!...
[ Full version ]

We introduce new dynamic set intersection data structures, which we call 2-3 cuckoo filters and hash tables. These structures differ from the standard cuckoo hash tables and cuckoo filters in that they choose two out of three locations to store each item, instead of one out of two, ensuring that any item in an intersection of two structures will have at least one common location in both structures. We demonstrate the utility of these structures by using them in improved algorithms...
[ Full version ]

This talk describes a system, known as PilotNet, that provides perception and control for self-driving cars. PilotNet learns from data, emulating the behavior of human drivers. The use of hand-crafted rules is kept to a minimum. As we gather more data we find that PilotNet performance keeps improving. Our test car, controlled by PilotNet, can drive, on average, over 20 miles on a highway before a human intervention is required. The core technology in PilotNet is convolutional neur...
[ Full version ]

A common way to speed up training of large convolutional networks is to add computational units. Training is then performed using data-parallel synchronous Stochastic Gradient Descent (SGD) with a mini-batch divided between computational units. With an increase in the number of nodes, the batch size grows. However, training with a large batch often results in lower model accuracy. We argue that the current recipe for large batch training (linear learning rate scaling with warm-u...
[ Full version ]

Jansen Huang, NVIDIA president and world techniology leader, will visit CS and will give a special talk: "AI WILL CHANGE THE WORLD". The lecture is open to all. ...
[ Full version ]

In 1979, Yao has defined the communication complexity model, including the two measures: Deterministic communication complexity and partition number. The relation between deterministic communication complexity and the partition number has been in the focus of much research in the field of communication complexity. Yao has noticed that the logarithm of the partition number is a lower bound on the deterministic communication complexity, and inquired about the exact relation. In...
[ Full version ]

Emerging large-scale applications rely on continuous tracking of complex queries over collections of massive, dynamic, and physically-distributed data streams. Thus, in addition to the space- and time-efficiency requirements of conventional stream processing (at each distributed site), effective solutions also need to guarantee communication efficiency. Continuously collecting the data to a central location is infeasible in large scale applications, as the excess communication r...
[ Full version ]

Element reduction in ultrasound transducers allows for several benefits, such asmore compact ultrasound systems and more efficient implementation of signal enhancement algorithms. However, element thinning poses two main challenges: SNR (signal to noise ratio) reduction and interference through grating lobes. Through the use of multiplicative beamforming in ultrasound systems, we decrease the number of processed elements in the transducer while preserving the beam pat...
[ Full version ]

We present the design and the implementation of IGATOOLS [1], a C++14 general purpose library for solving PDEs using the isogeometric analysis framework [2]. The most remarkable aspect of isogeometric methods is the use of the same set of spline functions for representing the geometric domain and for describing the solution of PDEs. In the IGATOOLS design, the mathematical concepts of the isogeometric method and their relationships are directly mapped into classes and t...
[ Full version ]

The majority of SAT solving applications are incremental in nature, i.e., instead of solving a single formula, a sequence of (syntactically) similar formulas are solved. In most cases the elements of the sequence are generated one after another at runtime, without a-priori knowledge of future instances. It is possible to accelerate the SAT search by using information gained from solving previous instances. Until recently, such information has been limited to sharing relevant learn...
[ Full version ]

Cloud computing has become ubiquitous in our lives. We store our Dropbox files and run our Google searches on the cloud. However, the cloud does not only provide storage and compute resources: it also provides significant networking resources that have largely gone unexplored in previous work. In my talk, I explore using public clouds to accelerate data transfers across the Internet. I aim to quantify the potential of cloudified data delivery and learn how to tap it. To...
[ Full version ]

We study the two-party communication and sample complexities of classification problems. We consider a variant of Yao's classical model in which the parties may transmit to each other examples from their input-samples rather than just bits. This enables (i) to investigate the notion of sample complexity — the total number of examples sent — which is a natural complexity measure in the context of learning theory, and (ii) to study infinite hypothesis ...
[ Full version ]

The Hiroshi Fujiwara Cyber Security Research Center will hold the 6th Summer School on Cyber and Computer Security: "Decentralized Cryptographic Currencies and Blockchains". The event will take place on Sunday-Thursday, September 10th-15th, 2017 at the Technion, Haifa. Tentative topics: Bitcoin, Ethereum and ZCash — overview Blockchain Scalability and Stability Game theory, economic incentives...
[ Full version ]

The traditional view of simulation is that processors are programmed and then simulators execute the processors threads. That is, the programmer thinks about processors when she creates the code. What if we tell her to think about simulators and write program for the simulator rather then the processors? Obviously, as we show, any processors' program is a simulator program and vice-versa. But thinking directly at simulators, results in, as we show, ``hig...
[ Full version ]

Modeling data has led to a revolution in the fields of signal and image processing, and machine learning. Consider the simplest restoration problem - removal of noise from an image. The recent advent of highly effective models for images (e.g. the sparse-land model) has led researchers to believe that existing denoisers are touching the ceiling in terms of restoration performance. Leveraging this impressive achievement, we propose a framework that is able to translate complicated ...
[ Full version ]

A secret sharing scheme allows a dealer to distribute a secret amongst a group of participants, where each participant is given a share of the secret. Authorized sets of participants can reconstruct the secret while unauthorized sets do not gain any information about the secret. We study the problem of sharing secrets in the presence of cheaters, namely parties who contribute incorrect shares to the reconstruction procedure. It is known that when there is a majority of cheaters,...
[ Full version ]

We introduce the concept of visibility detection within a point set to new domains. Specifically, we show that a simple representation of an image as a 3D point cloud lets us use visibility detection in classical image processing tasks, improving state-of-the-art results. Given an image, each pixel is represented as a feature point, a viewpoint is set, and points that are visible to the viewpoint are detected. What does it mean for a point to be visible? Although this question has...
[ Full version ]

We present a novel approach for automatic feature generation for predicting program properties. Our approach automatically produces features that can capture long-distance syntactic relationships between program elements. The features are purely syntactic, and the method is useful for any programming language. Inspired by Parse Tree Paths in Natural Language Processing (NLP), we generate features that capture relationships in an Abstract Syntax Tree (AST). We show that these f...
[ Full version ]

The talk includes two parts. First, we consider the problem of designing repair-efficient distributed storage systems that are secure against a passive eavesdropper that can observe the contents of any subset of nodes of limited size. We present a universal framework for the design of coset-coding based outer codes that enables information-theoretic security properties for a broad class of storage codes. As case studies, we consider minimum storage regenerating codes with small r...
[ Full version ]

The Total Variation (TV) framework has been shown to give a good scale-space representation for many purposes. Much has been done in showing the uses of TV for denoising, deconvolution and other spectral analysis tasks both for one-dimensional signals and for images, mainly for amplitude and grayscale images. In the field of colour imaging and compression, it is known that chromatic channels should be sampled and transmitted at a lower rate than that of the intensity. This is used...
[ Full version ]

We consider a group of mobile agents on a line, identical and indistinguishable, memoryless, having the capability to only sense the presence of neighboring agents to the left and to the right. The agents' rule of motion is as follows : at each moment, agents with neighbors on both sides stay put, while agents with neighbors on one side only jump with high probability a unit distance towards the neighbors (otherwise, they jump one unit away). We prove that all agents, ex...
[ Full version ]

Semantic parsing is the task of mapping natural language sentences into a formal representation of their meaning, often defined as logical forms. One of the prominent uses of semantic parsing is parsing natural language instructions in the context of natural language interfaces (NLIs) for various types of software applications. In this work, we present a novel task: parsing instructions in simple domains that are unseen during training, into logical forms with deep compositionalit...
[ Full version ]

Textures are what differentiates true, real-life images, from cartoon images. The latter emphasize mainly smooth content other than the edges and contours, whereas the former stresses the importance of the details within the contours. Textures are found in facial images, natural scenery, aerial, medical and other types of images, and affect the image perception and recognition. Images with smoothed-out textures appear artificial and cartoon-like. This study is devoted...
[ Full version ]

We address the task of Named Entity Disambiguation (NED) for noisy text. We present WikilinksNED, a large-scale NED dataset of text fragments from the web, which is significantly noisier and more challenging than existing news-based datasets. To capture the limited and noisy local context surrounding each mention, we design a neural model based on GRUs and attention and describe how to train it. We also describe a new way of initializing word and entity embeddings that significa...
[ Full version ]

We present an interactive tree modeling and deformation system that supports an efficient collision detection and avoidance using a bounding volume hierarchy of sweep surfaces. Starting with conventional tree models (given as meshes), we convert them into sweep surfaces and deform their branches interactively while detecting and avoiding collisions with many other branches. Multiple tree models (sharing the same topology) can be generated with great ease using the swe...
[ Full version ]

Mutually Uncorrelated (MU) codes are a class of codes in which no proper prefix of one codeword is a suffix of another codeword. These codes were originally studied for synchronization purposes and recently, Yazdi et al. showed their applicability to enable random access in DNA storage. In this work we follow the research of Yazdi et al. and study MU codes along with their extensions to correct errors and balanced codes. We first review a well-known construction of MU codes and s...
[ Full version ]

Convolutional neural networks (CNNs) provide the current state of the art in visual object classification, but they are far less accurate when classifying partially occluded objects. A straightforward way to improve classification under occlusion conditions is to train the classifier using partially occluded object examples. However, training the network on many combinations of object instances and occlusions may be computationally expensive. This work proposes an alternative appr...
[ Full version ]

Axis-parallel tests are variants of low-degree tests where the input function can only be examined via its restrictions to axis-parallel lines or hyperplanes. In this talk, we present two new results on axis-parallel tests. (1) We prove the first analogue of the Bivariate Low-Degree Testing Theorem of Polishchuk and Spielman (1994) that works for arbitrarily small agreement. Unlike previous works, our proof techniques are combinatorial, not algebraic. (2) We im...
[ Full version ]

Existing constructions of zero knowledge proof systems typically work as follows: construct a proof system which is information-theoretically sound but not zero knowledge, then apply some cryptographic transformation to obtain a proof or argument system which is zero-knowledge. In this talk I will suggest a different approach: construct a proof system (in an extended model) which is information-theoretically sound /and/ zero knowledge, and then apply a ZK-preserving cryptographic ...
[ Full version ]

As core counts continue to rise in modern processors, it is increasingly important for applications to be scalable. Designing scalable concurrent software is notoriously difficult (even for experts), and programmers must rely on efficient concurrent library code to be effective. Concurrent binary search trees (BSTs) represent some of the most fundamental building blocks in such libraries, and are important in both theory and practice. This talk will briefly introduce two techni...
[ Full version ]

Flash memories is a non-volatile technology that is both electrically programmable and electrically erasable. It incorporates a set of cells maintained at a set of levels of charge to encode information. While raising the charge level of a cell is an easy operation, reducing the charge level requires the erasure of the whole block to which the cell belongs, which is a slow operation and damages the life-time of the device. I will present some of our results regarding two codin...
[ Full version ]

The quest for algorithms which enable cognitive abilities is an important part of machine learning. A common trait in these recent cognitive-like tasks is that they take into account different data modalities, e.g., visual and lingual. We propose a novel and generally applicable form of attention mechanism that learns high-order correlations between various data modalities. We show that high-order correlations effectively direct the appropriate attention to the relevant elements i...
[ Full version ]

Have you ever wondered about “order amidst chaos” in a “crazy” intersection where drivers, bikers and pedestrians manage to cross it without collisions?​ This talk is about low level visual cues that can help to explain the sensing and actions of different non-colliding moving vehicles and people. The talk focuses on one of the cues, called “Visual Looming” that can be measured and act-upon locally at the moving agent level.​ ​In this presentation we will explain...
[ Full version ]

In this talk, we will discuss the classic bipartite matching problem in the online setting, first introduced in the seminal work of Karp, Vazirani and Vazirani (STOC '90). Specifically, we consider the problem for the well-studied class of regular graphs. Matching in this class of graphs was studied extensively in the offline setting. In the online setting, an optimal deterministic algorithm, as well as efficient algorithms under stochastic input assumptions were known. In this wo...
[ Full version ]

IBM recently released IBM Quantum Experience, a freely accessible programming interface to its 5-bit universal quantum computer, accompanied by a 16-bit beta version, both set on the IBM Cloud. I will describe the science and technology behind this milestone event. I will then present IBM's further vision for quantum computing in the marketplace, in terms of hardware, software stack, potential applications, and algorithms. Short Bio: ========== Yehuda Naveh got his B.Sc. ...
[ Full version ]

On Tuesday, July 4th, 2017, the Systems and Software Development Laboratory (SSDL) will hold a project Fair on IoT and Android, presenting the newest and most inspiring projects presented by the developing teams. You are all invited! Following are the presenting projects: IOT BreakFast Have your favorite breakfast ready when you wake up (...
[ Full version ]

Theory graduate students will present their field of study alongside some of the problems they are conducting research on. More details in the attached poster. Everyone is welcome to attend, no special background is assumed. Refreshments will be served. ...
[ Full version ]

We consider a discrete system of "ant-like" agents which pursue each other on the vertices of a graph environment. Visually reminiscent of a trail of ants, the agents emerge one by one at equal time intervals from a source vertex s and pursue each other by greedily attempting to close the distance to their immediate predecessor, the agent that emerged just before them from s, until they arrive at the destination point t. Such pursuits have been investigated before in the continuou...
[ Full version ]

Sketch-based streaming algorithms allow efficient processing of big data. These algorithms use small fixed-size storage to store a summary ("sketch") of the input data, and use probabilistic algorithms to accurately estimate the desired quantity. A fundamental streaming problem is "cardinality estimation": given a very long stream of elements, the goal is to estimate the number of distinct elements. This is a well-known problem with numerous applications for network monitor...
[ Full version ]

Machine virtualization has grown in popularity in recent years with the growth of cloud computing. Virtual machines use virtual I/O devices to perform their I/O. Nowadays paravirtual I/O devices are the most popular type of virtual I/O devices due to their high performance and interposition capabilities. However paravirtual I/O devices also have disadvantages. Users need to install device drivers for paravirtual devices whenever they switch hypervisors, and hypervisor providers ne...
[ Full version ]

The Geometric Image Processing Laboratory (GIP) and the Center for Graphics and Geometric Computing (CGGC) invite you to a special event of exposure and early registration to their joint course: Virtual and Augmented Reality. The event will be held as part of the End of Year party, on Wednesday, June 28, 2017, between 12:30-14:30at the CS Taub Lobby. More deta...
[ Full version ]

The Internet's communication infrastructure (TCP/IP, DNS, BGP, etc.) is alarmingly insecure, as evidenced by many high-profile incidents. I will illustrate the challenges en route to securing the Internet, and how these can be overcome, by focusing on the Internet's, arguably, biggest security hole: the vulnerability of Internet routing to traffic hijacking attacks. Bio: Michael Schapira is an associate professor at the School of Computer Science and Engineering...
[ Full version ]

Graphical models, including constraint networks, Bayesian networks, Markov random fields and influence diagrams, have become a central paradigm for knowledge representation and reasoning in Artificial Intelligence, and provide powerful tools for solving problems in a variety of application domains, including coding and information theory, signal and image processing, data mining, learning, computational biology, and computer vision. Although past decades have seen considerable pro...
[ Full version ]

Sparse-to-dense interpolation for optical flow is a fundamental phase in the pipeline of most of the leading optical flow estimation algorithms. The current state-of-the-art method for interpolation, EpicFlow, is a local average method based on an edge aware geodesic distance. We propose a new data-driven sparse-to-dense interpolation algorithm based on a fully convolutional network. We draw inspiration from the filling-in process in the visual cortex and introduce lateral depende...
[ Full version ]

We propose a novel technique for computing consistent cross fields on a pair of triangle meshes given an input correspondence, which we use as guiding fields for approximately consistent quadrangulations. Unlike the majority of existing methods our approach does not assume that the meshes share the same connectivity or even have the same number of vertices, and furthermore does not place any restrictions on the topology (genus) of the shapes. Importantly, our method is robust with...
[ Full version ]

By combining geometric principles for shape analysis with modern sensing techniques, large-scale datasets and powerful computational architectures we show various new ways of enabling computers to better perceive, interpret and comprehend the geometry of the world around them. Specifically we explore the topics of reconstruction, filtering and semantic processing within the context of computational geometric vision. Reconstruction We start by discussing the problem of sensing...
[ Full version ]

Most communication networks exhibit a large variety of links that may differ significantly in terms of the cost incurred for sending messages over them. While this fact is reflected in many centralized network optimization problems, it is not yet taken into account in the design of distributed algorithms: The de facto standard approach for measuring the communication burden of distributed message passing algorithms is to minimize their message complexity — the total number of me...
[ Full version ]

The interest in acquiring and analyzing the geometry of the world is ever increasing, fueling a wide range of computer vision algorithms in the field of geometry processing. Spectral analysis has become key component in many applications involving non-rigid shapes modeled as two-dimensional surfaces, and recently, convolutional neural networks have shown remarkable success in a variety of computer vision tasks. We designed a set of methods and tools that use these paradigms fo...
[ Full version ]

The 7th annual international TCE conference will take place on Wednesday-Thursday, June 21-22, 2017 at the Technion CS Taub Building and will focus onadvancing information technologies like distributed storage and cloud systems, new and emerging memory technologies, biological systems and genomics, secrecy and security, reliable computation and big data, communications and networking. Conference Chairs: Eitan Yaakobi (CS Technion) and Yuval Cassuto (EE Techni...
[ Full version ]

The end of CMOS transistors scaling marks a new era for modern computer systems. As the gains from traditional general-purpose processors diminish, researchers explore the new avenues of domain-specific computing. The premise of domain-specific computing is that by co-optimizing software and specialized hardware accelerators, it is possible to achieve higher performance per power rates. In contrast to technology scaling, specialization gains are not sustainably scalable, as there ...
[ Full version ]

I'll present two new concepts in deep learning and show how we used them to achieve a significant improvement in stereo matching, which is one of the most fundamental problems in computer vision. The first is a new residual architecture dedicated for metric learning, and the second is a general way to assess the confidence in the network's prediction. *Amit is a deep learning and computer vision engineer at Magic Leap and M.Sc student in deep learning and computer...
[ Full version ]

Stochastic Gradient Descent (SGD) is a popular method for learning classes of linear predictors. If the size of the training set is large, a computational bottleneck in SGD is the computation of the gradient, and hence, it is common to distribute the gradient computation among worker nodes. However, in distributed computation scenarios, stragglers (i.e., slow or unavailable nodes) might cause a considerable delay, and hence, schemes for mitigation of stragglers are essential. ...
[ Full version ]

Title: GWCNN: A Metric Alignment Layer for Deep Shape Analysis Abstract: Deep neural networks provide a promising tool for incorporating semantic information in geometry processing applications. Unlike image and video processing, however, geometry processing requires handling unstructured geometric data, and thus data representation becomes an important challenge in this framework. Existing approaches tackle this challenge by converting point clouds, meshes, or ...
[ Full version ]

The 2017 Workshop in Cryptology will be held on Thursday, June 15 2017, between 9:00-17:15, in Auditorium 1, at the CS Taub Building, Technion. Most talks will be held in Hebrew and Keynote speaker will be Dr. Marc Stevens who will talk about his recent work that found a collision of SHA-1. Other highlights will be talks about virtual coins and blockchains, and other topics. More ...
[ Full version ]

Which features make the brain such a powerful and energy-efficient computing machine? Can we reproduce them in the solid state, and if so, what type of computing paradigm would we obtain? I will show that a machine that uses memory to both process and store information, like our brain, and is endowed with intrinsic parallelism and information overhead - namely takes advantage, via its collective state, of the network topology related to the problem - has a computational power ...
[ Full version ]

Distributed and cloud storage systems today are required to tolerate the failure or unavailability of some of the nodes in the system. The simplest and most commonly used way to accomplish this task is replication, whereby every node is replicated several times, usually three. This solution has clear advantages due to its simplicity, fast recovery, and efficient availability. However, it entails a large storage overhead which becomes costly in large storage systems. In this wo...
[ Full version ]

More details in the attached ad. You are all invited....
[ Full version ]

The cryptographic hash function SHA-1 has been known to be weak since 2004 with the first theoretical collision attack. Since then many techniques for SHA-1 collision attacks have been developed and refined over the years leading to our recent practical collision attack. In this tutorial I will present the techniques underlying our attack in detail covering things such as local collisions, disturbance vectors, differential trail construction, computing optimal...
[ Full version ]

I will describe several old and new applications of topological and algebraic methods in the derivation of combinatorial results. In all of them the proofs provide no efficient procedures for solving the corresponding algorithmic questions. The problem of finding such procedures (or convincing reasons indicating that they are unlikely to exist) is an intriguing challenge. ...
[ Full version ]

Safety of a distributed protocol means that the protocol never reaches a bad state, e.g., a state where two nodes become leaders in a leader-election protocol. Proving safety is obviously undecidable since such protocols are run by an unbounded number of nodes, and their safety needs to be established for any number of nodes. I will describe a deductive approach for proving safety, based on the concept of universally quantified inductive invariants --- an adaptation of the mathema...
[ Full version ]

In the talk, I will present a distillation algorithm which operates on a large, unstructured, and noisy collections of internet images returned from an online object query. I will introduce the notion of a distilled set, which is a clean, coherent, and structured subset of inlier images, where the object of interest is properly segmented out throughout the set. I will also demonstrate the utility of our technique with a number of interesting graphics applications, including a nove...
[ Full version ]

Mutually Uncorrelated (MU) codes are a class of codes in which no proper prefix of one codeword is a suffix of another codeword. These codes were originally studied for synchronization purposes and recently, Yazdi et al. showed their applicability to enable random access in DNA storage. In this work we follow the research of Yazdi et al. and study MU codes along with their extensions to correct errors and balanced codes. We first review a well-known construction of MU codes and s...
[ Full version ]

We will discuss a variation of the Ginzburg-Landau functional, a common tool in applications such as image segmentation (Ambrosio-Tortorelli) and phase-field methods in fluid simulation, involving a so-called "double-obstacle" barrier term (first studied by Elliott and Blowey). We will describe fast (GPU-optimized) variational solvers for gradient flows of these functionals (Allen-Cahn and Cahn-Hilliard equivalents), and also look into certain higher-dimensional genera...
[ Full version ]

Multidimensional NMR (MDNMR) experiments are an important tool in physical chemistry,but can take a long time, in some cases weeks, to conduct. At first glance, the application looks ideal for compressed sensing because the object to be recovered is sparse and the under-sampled measurements are made in the 'Fourier' domain. Actually, MDNMR is not covered by the existing compressed sensing literature. First the 'Fourier' domain is not the classical one, but involves the so-...
[ Full version ]

To prove space lower bounds​ for refuting random k-CNF formulas in the UNSAT region we use the expansion property of the incidence graph of the formula, which guarantees large matchings. In Polynomial Calculus (PC), a proof system working with polynomials, simple matchings are no longer sufficient to obtain good space bounds and we need to guarantee large bi-matchings (or V-matchings). Nevertheless, for 3-CNFs even bi-matchings are not longer sufficient. ...
[ Full version ]

With the rapid growth in machine-generated traffic and in storage capacities offered by Web mail services, our mailboxes get larger and larger. Search therefore becomes a critical mechanism in order to retrieve the specific messages we need. Unfortunately, during the past decade mail search has not observed the same pace of progress as Web search and is still perceived as a difficult and frustrating task. Only recently, mail search has become an emerging topic for research and has...
[ Full version ]

Solid State Disks (SSDs) have the potential to revolutionize the storage system landscape. They have gained popularity as cache devices in data centers because, they are faster in read and write operations and have lower power consumption, compared to the traditional magnetic hard disks (HDD). However, SSDs have a limited number of times it can write to a single physical location, and there is a second limitation. The SSD must perform an erase operation before it can write to the...
[ Full version ]

One of the most common algorithmic tasks is to find a single interesting event (a needle) in an exponentially large collection (haystack) of N=3D2^n possible events, or to demonstrate that no such event is likely to exist. In particular, we are interested in the problem of finding needles which are defined as events that happen with an unusually high probability of p>>1/N in a haystack which is an almost uniform distribution on N possible events. Such a search algori...
[ Full version ]

I will present our pipeline for optical flow computing, based on a CNN for generating local descriptors. I will focus on our recent research, where we show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric l...
[ Full version ]

You are invited to the upcoming 2017 Stephen and Sharon Seiden Frontiers in Engineering and Science Workshop. This year, the workshop will be titled "Beyond CMOS: From Devices to Systems" and will be held at the Technion, Haifa, Israel on Monday-Tuesday, June 5-6, 2017. This workshop will bring together researchers and leaders from academia and industry to discuss the many different aspects of emerging solid state memories including device phy...
[ Full version ]

Lattices are error-correcting codes defined over the real numbers. In the physical world, 1 plus 1 is 2, and it is the same for lattices. An important example of a physical-world application is wireless communications. Two electromagnetic signals, transmitted at the same time, will superimpose --- that is, they add, making lattice codes a natural fit for wireless communications. This lecture gives a brief introduction to lattices for those already familiar with the fundamentals...
[ Full version ]

When visualising topological objects via 3D printing, we need a three-dimensional geometric representation of the object. There are approximately three broad strategies for doing this: "Manual" - using whatever design software is available to build the object by hand; "Parametric/Implicit" - generating the desired geometry using a parametrisation or implicit description of the object; and "Iterative" - numerically solving an optimisation problem. The manual strategy is...
[ Full version ]

This work proposes an algorithm for computing dense packings of congruent circles inside general 2D containers. Unlike the previous approaches which accept as containers, only simple, symmetric shapes such as circles, rectangles and triangles, our method works for any container with a general, freeform (spline) boundary. In contrast to most previous approaches which cast the problem into a non-convex optimization problem, our method attempts to maximize the number of ...
[ Full version ]

The lecture will be given as part of the Network Function Virtualization advanced course (cs236635). Dudu Amzallag is Technion CS Ph.D. graduate and is Group Head of Network Virtualisation, SDN and NFV, Vodafone. ...
[ Full version ]

Motivated by systems where the information is represented by a graph, such as neural networks, associative memories, and distributed systems, we present in this work a new class of codes, called codes over graphs. Under this paradigm, the information is stored on the edges of an undirected graph, and a code over graphs is a set of graphs. A node failure is the event where all edges in the neighborhood of the failed node have been erased. We say that a code over graphs can tolerate...
[ Full version ]

The mainstream approaches in digital geometry processing utilize triangular (simplicial) meshes, discretize differential quantities using finite-element function spaces, and describe transformations with piecewise affine maps. I will describe how Moebius geometry provides an original alternative to discrete differential geometry, by using circles as its basic elements, and describing quantities like conformality and regularity through the invariant cross-ratio. This theory allows ...
[ Full version ]

Elbit Aerospace utilizes 3D mapping in a variety of different projects. We present two of these projects; in the first project, a 3D model of a large area is generated from aerial imaging. Here, the rapidly changing viewpoint of an aircraft mounted camera enables multi view geometry techniques to obtain a 3D point cloud. A mesh is reconstructed, and images from the same camera are used to create a photorealistic model. The composed model is 2.5D, that is, it includes one height f...
[ Full version ]

Advances in machine learning, notably deep learning, have led to computers matching or surpassing human performance in several cognitive tasks including vision, speech and natural language processing. However, implementation of such neural algorithms in conventional "von-Neumann" architectures are several orders of magnitude more area and power expensive than the biological brain. Exploring the new paradigm of computing necessitates a multi-disciplinary approach: exploration of ne...
[ Full version ]

This talk describes the importance technology plays in financial services. Amir will describe several examples of domains where technology is disrupting the market, together with the challenges our SW engineers face on a day to day basis in building these technologies. No background of finance is assumed or required. Bio: Amir Nahir received his Bsc, Msc and Phd from the Technion in 2006, 2008 and 2014 (respectively). He currently leads the development of business-...
[ Full version ]

We present a sublinear-time algorithm for approximating the number of $k$-cliques in an input graph. We assume the standard general graphs access model via (1) degree queries, (2) neighbor queries and (3) pair queries. Consider a graph with $n$ vertices, $m$ edges, and $k$-cliques. We design an algorithm that outputs a $(1+\eps)$-approximation (with high probability) for $k$ with expected query complexity and running time $O\left(\frac{n}{\clk^{1/k}}+\frac{m^...
[ Full version ]

Sparse data computations are ubiquitous in science and engineering. Two widely used applications requiring sparse data computations are graph algorithms and linear algebra operations such as Sparse Matrix-Vector Multiplication (SpMV). In contrast to their dense data counterparts, sparse-data computations have less locality and more irregularity in their execution - making them significantly more challenging to optimize. While there are several existing formats for sparse data repr...
[ Full version ]

Verification of concurrent programs is known to be extremely difficult. On top of the challenges inherent in verifying sequential programs, it adds the need to consider a high (typically unbounded) number of thread interleavings. In this work, we utilize the plethora of work on verification of sequential programs for the purpose of verifying concurrent programs. We introduce a technique which reduces the verification of a concurrent program to a series of verification tasks ...
[ Full version ]

The driving force behind convolutional networks - the most successful deep learning architecture to date, is their expressive power. Despite its wide acceptance and vast empirical evidence, formal analyses supporting this belief are scarce. The primary notions for formally reasoning about expressiveness are efficiency and inductive bias. Efficiency refers to the ability of a network architecture to realize functions that require an alternative architecture to be much larger. Induc...
[ Full version ]

The seventh CS Research Day for graduate studies will be held on Tuesday, May 21, 2017, between 15:30-176:30, at the lobby of the CS Taub Building. Research Day events are opportunity for our graduate students to expose their researches using posters and presentations to CS faculty and all degrees students, Technion distinguished representatives and to high-ranking delegates from the hi-tech leading industry companies in Israel and abroad. The partic...
[ Full version ]

Four unsolved problems that originate from research in Computer Graphics and Geometric Modeling will be presented. The first problem involves understanding the notion oscillation for Bezier surfaces, the freeform polynomial surfaces most common in Computer Graphics and Geometric Modeling. The second problem concerns gnerating smooth (C2) surfaces via subdivision from triangular or quadrilateral meshes of arbitrary topology. The third problem ...
[ Full version ]

Racetrack memory is a new technology which utilizes magnetic domains along a nanoscopic wire in order to obtain extremely high storage density. In racetrack memory, each magnetic domain can store a single bit of information, which can be sensed by a reading port (head). The memory has a tape-like structure which supports a shift operation that moves the domains to be read sequentially by the head. In order to increase the memory's speed, prior work studied how to minimize the late...
[ Full version ]

Piecewise polynomial constraint systems are common in numerous problems in computational geometry, such as constraint programming, modeling, and kinematics. In this talk, we present a framework that is capable of decomposing, and efficiently solving a wide variety of complex piecewise polynomial constraint systems. The framework we present uses a constraint system decomposition algorithm to break down complex problems into smaller, simpler subproblems. It ...
[ Full version ]

The classical Sylvester-Gallai theorem states the following: Given a finite set of points in the Euclidean plane, if the line through every pair of points passes through a third point, then all points must be collinear. Thus basically the result shows that many `local' linear dependencies implies a `global' bound on the dimension of the entire set of points. Variations of these questions have been well studied in additive combinatorics and incidence geometry. In the last few year...
[ Full version ]

Big data is revolutionizing the way we live, work, and socialize. This revolution is powered by datacenters built with commodity hardware technologies that are now running out of steam. In this talk, I will outline the disruptive challenges facing future computing systems and will argue for specializing server hardware as the only way forward. The principal challenge for specialization lies in providing common-case efficiency benefits without losing the generality necessary to acc...
[ Full version ]

While the jury is still out as to whether the impressive experimental progress on quantum gates and qubits will lead one day to a full scale quantum computing machine, a new and exciting development has been taking place over the past decade. Computational notions such as reductions, hardness, and completeness are quickly starting to be integrated into the very heart of the research of many body quantum systems. The computational perspective brings deep new insights in...
[ Full version ]

Ultrasound images are often contaminated with acoustic clutter, which obscures image details of interest, thus leading to potentially inaccurate medical diagnosis. In order to address this problem, we are proposing a model-based image reconstruction approach using the individually stored channel data of the ultrasound transducer elements, in a grid of image points. Analysis of the data allows the description of the image as consisting of coherent strong reflectors, speckled tissue...
[ Full version ]

Motivated by the error behavior in DNA storage channels, in this work we extend the previously studied sequence reconstruction problem by Levenshtein. The reconstruction problem studies the model in which the information is read through multiple noisy channels, and the decoder, which receives all channel estimations, is required to decode the information. For the combinatorial setup, the assumption is that all the channels cause at most t errors. ...
[ Full version ]

We introduce the notion of depth with respect to a finite family F of convex sets in R^d that generalizes the well-studied Tukey depth. Specifically, we say that a point p has depth m with respect to F if every hyperplane that contains p intersects at least m sets of F. We study some nice properties of Tukey depth that extend to this definition and point out some key differences. By imposing additional intersection hypothesis to the family F, we prove a centerpoint...
[ Full version ]

Recent years has seen an explosion in the number of connected devices, which means not only growth in velocity and volume of data, but also that data sources are increasingly geographically distributed, raising cost of communication. Data mining algorithms often assume that data is centralized or that communication is inexpensive: the setting is implicitly assumed to be a data center. In settings like wireless sensor networks, however, communication costs battery power. Moreover,...
[ Full version ]

Abstract We show that there exist binary locally testable codes (for all rates) and locally correctable codes (for low rates) with rate and distance approaching the Gilbert-Varshamov bound (which is the best rate-distance tradeoff known for general binary error-correcting codes). Our constructions use a number of ingredients: Thommesen's random concatenation technique, the Guruswami-Sudan-Indyk strategy for list-decoding concatenated codes, the Alon-Edmonds-Luby distance amplifica...
[ Full version ]

In the last decade, Web mail traffic has evolved, very much like regular snail mail, into being dominated by machine- generated messages. Some recent studies have verified that more than 90% of non-spam Web email is indeed generated by automated scripts. Although generated by machines, a large part of these messages include highly personal information, e.g. bank statements, travel plans, or shipment notifications. In this presentation, we will provide an overview on how machine ge...
[ Full version ]

Over the years, a few programming paradigms have been successful enough to enter the casual vocabulary of software engineers: procedural, imperative, object-oriented, functional, generic, declarative. There's a B-list, too, that includes paradigms such as logic, constraint-oriented, and symbolic. The point is, there aren't very many of them altogether. Easy to imagine, then, the immensely humbling pressure one must feel when stumbling upon a way to think about writing code...
[ Full version ]

In this work we study two families of codes with availability, namely \emph{PIR codes} and \emph{batch codes}. While the former requires that every information symbol has $k$ mutually disjoint recovering sets, the latter asks this property for every multiset request of $k$ information symbols. The main problem under this paradigm is to minimize the number of redundancy symbols. We denote this value by $r_P(n,k), r_B(n,k)$, for PIR, batch codes, respectively, where $n$ is the numbe...
[ Full version ]

Facial expressions play a significant role in human communication and behavior.​ ​Psychologists have long studied the relationship between facial expressions and emotions.​ ​Paul Ekman et al., devised the Facial Action Coding System (FACS) to taxonomize human facial expressions and model their behavior.​ ​The ability to recognize facial expressions automatically, enables novel applications​ ​in fields like human-computer int​​eraction, social gaming, and psych...
[ Full version ]

A spanner of a given graph is a sparse subgraph that approximately preserves distances. Since their introduction in the late 1980's, spanners have found numerous applications in synchronization problems, information dissemination, routing schemes and more. Many applications of spanners are in computer networks, where the network needs to find a spanner for its own communication graph. We present distributed algorithms for constructing additive spanners in networks of bounded m...
[ Full version ]

The rapid development of memory technologies has introduced challenges to the continued scaling of memory devices in density and access speed. Many of these challenges can be mitigated by coding techniques, which optimize the representation of data within these memories. In this talk, we present a practically-motivated model, where the use of novel coding frameworks improves performance. We start with introducing the class of partial-erasure channels, where the channel output is a...
[ Full version ]

In this talk, I will briefly introduce myself, mainly focusing on my doctoral dissertation, addressing realistic facial animation. Realistic facial synthesis is one of the most fundamental problems in computer graphics, and is desired in a wide variety of fields, such as film and advertising, computer games, teleconferencing, user-interface agents and avatars, and facial surgery planning. In the dissertation, we present the most commonly pract...
[ Full version ]

I will describe the notion of agreement testing, which allows to deduce global structure from local agreement checks. In retrospect, agreement testing theorems are a key combinatorial component in nearly all PCPs. I will describe a couple of recent works. The first shows how an agreement testing theorem (which we don't yet know how to prove) would imply NP hardness for unique games with gap of 1/2-ε​ vs ε​. The second is a new agreement testi...
[ Full version ]

The 2016 open day invites outstanding undergraduates from all universities to come to the Technion and learn about the faculties of Computer Science and  Electrical Engineering, to meet faculty and graduate students and to hear a fascinating talk by Prof. Lior Kornblum: "How to Connect Exotic Physics with Future Devices" and by Oded Cohen, VP in IBM Haifa: "Do Advanced Degrees Indeed Advance?" The event will b...
[ Full version ]

In this talk I will show how ''blind'' visual inference can be performed by exploiting the internal redundancy inside a single visual datum (whether an image or a video). The strong recurrence of patches inside a single image/video provides a powerful data-specific prior for solving complex tasks in a ''blind'' manner. The term ''blind'' here is used with a double meaning: (i) Blind in the sense that we can make sophisticated inferences about things we have never seen...
[ Full version ]

Many shape analysis methods treat the geometry of an object as a metric space captured by the Laplace-Beltrami operator. We present an adaptation of a classical operator from quantum mechanics to shape analysis where we suggest to integrate a scalar function through a unified elliptical Hamiltonian operator. We study the addition of a potential function to the Laplacian as a generator for dual spaces in which shape processing is performed. After exploration of the decomposition o...
[ Full version ]

We investigate the asymptotic rates of length-$n$ binary codes with VC-dimension at most $dn$ and minimum distance at least $\delta n$. Two upper bounds are obtained, one as a simple corollary of a result by Haussler and the other via a shortening approach combining Sauer--Shelah lemma and the linear programming bound. Two lower bounds are given using Gilbert--Varshamov type arguments over constant-weight and Markov-type sets. ...
[ Full version ]

General-Purpose (GP) CPUs are a quintessential example for “engineering for serendipity” as their current ubiquity seems to exceed the wildest dreams of its originators. Alas, in 2017, the only off-the-shelf GPCPU is a single core performing only marginally better since 2004, when an era of phenomenal GP performance growth ended. Furthermore, once technology constraints forced a transition to many-cores, CPU designers chose to volunteer general programmers to help achieve per...
[ Full version ]

We propose a design for a fine-grained lock-based skiplist optimized for Graphics Processing Units (GPUs). While GPUs are often used to accelerate streaming parallel computations, it remains a significant challenge to efficiently offload concurrent computations with more complicated data- irregular access and fine-grained synchronization. Natural building blocks for such computations would be concurrent data structures, such as skiplists, which are widely used in general purpo...
[ Full version ]

We present Omid - a low latency transaction processing service that powers web-scale production systems at Yahoo. Omid provides ACID transaction semantics on top of traditional key-value storage; its implementation over Apache HBase is open sourced as part of Apache Incubator. Omid can serve hundreds of thousands of transactions per second on standard mid-range hardware, while incurring minimal impact on the speed of data access in the underlying key-value store. Additionally, as ...
[ Full version ]

The counter-intuitive features of Quantum Mechanics make it possible to solve problems and perform tasks that are beyond the abilities of classical computers and classical communication devices. Entanglement is an important feature of quantum states, and it is important in quantum information, quantum communication, and quantum computing. In the main part of this talk, we provide a geometrical analysis of entanglement and separability for all the quantum mixed states that are of ...
[ Full version ]

There are various theoretically-efficient constructions of public-randomness (i.e., Arthur-Merlin type) Zero-Knowledge Succinct Arguments of Knowledge in the random oracle model, for computations-verification (also known as verifiable-computation and computational-integrity). Those constructions could be used to solve many real world problems; Unfortunately, reducing those fabulous theoretical systems to broadly-adaptable implementations is a challenging task. This talk surveys ou...
[ Full version ]

The ability to detect similar actions across videos can be very useful for real-world applications in many fields. In this talk I will describe the descriptor we developed, the "temporal-needle" descriptor. Our descriptor captures the dynamic behavior, while being invariant to viewpoint and appearance. Using the descriptor, we were able to detect the same behavior across videos in a variety of scenarios. I will describe how the descriptor is computed, and how it can be used to fin...
[ Full version ]

We present an analysis method to improve treatment procedures for a set of neurological pathologies, specifically essential tremor disorder. By combining anatomical and pathological knowledge with modern geometric analysis tools of MRI, an efficient focused ultrasound treatment procedure is introduced. As a first step, we apply statistical geometric tools for in-vivo analysis of the brain. The brain's connectivity structure is interpreted based on diffusion tensor-MRI. Next...
[ Full version ]

Direct network I/O allows network controllers (NICs) to expose multiple instances of themselves, to be used by untrusted software without a trusted intermediary. Direct I/O thus frees researchers from legacy software, fueling studies that innovate in multitenant setups. Such studies, however, overwhelmingly ignore one serious problem: direct memory accesses (DMAs) of NICs disallow page faults, forcing systems to either pin entire address spaces to physical memory and thereby hinde...
[ Full version ]

The premise of the talk is processing natural language using machine learning techniques. While deep learning methods in Natural Language Processing are arguably overhyped, recurrent neural networks (RNNs), and in particular LSTM networks, emerge as very capable learners for sequential data. Thus, my group started using them everywhere. After briefly explaining what they are and why they are cool, I will describe some recent work in w...
[ Full version ]

While augmenting a system with multiple GPUs is an appealing way to push more compute power inside a single machine, it is not without its challenges. We will talk about these challenges in the context of network servers, where short request handling latency and throughput scaling with the number of GPUs are the main design goals. We claim that the current GPU programming model, where the GPU is a co-processor and is managed by the CPU, is an impediment to building sca...
[ Full version ]

The theory of software creation (aka programming) relies on several theoretical domains: Computability, complexity, algorithms, graph theory, and alike. Alas, the above are related to only a narrow facet of the quality of the software - functional correctness and performance. When software is created automatically, correctness and performance may be enough. Until then, two decades or more into the future, quality software is also measured by usability, testability, an...
[ Full version ]

Net neutrality is getting major attention these days, as it is at the crossroads between technology, economics and regulation. Discussions and new net neutrality rules and remedies are commonplace, all as a part of the attempt to find a balance between the need to follow common-carriage principles and the desire to provide different QoS by selectively blocking, slowing or providing faster connection tracks to web traffic of different customers with different applications by the IS...
[ Full version ]

1. Freehand Laser Scanning Using Mobile Phone 3D scanners are growing in their popularity as many new applications and products are becoming a commodity. These applications are often tethered to a computer and/or require expensive and specialized hardware. In this chapter of the thesis we demonstrate that it is possible to achieve good 3D reconstruction on a mobile device. We describe a novel approach for mobile phone scanning which utilizes a smart-phone and cheap laser poin...
[ Full version ]

Supervised and unsupervised methods have been used extensively to analyze genomics data, with mixed results. On one hand, new insights have led to new biological findings. On the other hand, analysis results were often not robust. Here we take a look at several such challenges from the perspectives of networks and big data. Specifically, we ask if and how the added information from a biological network helps in these challenges. We show both examples where the network added i...
[ Full version ]

A coreset (or core-set) of a dataset is its semantic compression with respect to a set of queries, such that querying the (small) coreset provably yields an approximate answer to querying the original (full) dataset. However, we are not aware of real-time systems that compute coresets in a rate of dozens of frames per second. I will suggest a framework to turn theorems to such systems using coresets. This is by maintaining such a coreset for kinematic (moving) set of n points, and...
[ Full version ]

Erasure codes protect data in large scale data centers against multiple concurrent failures. However, in the frequent case of a single node failure, the amount of data that must be read for recovery can be an order of magnitude larger than the amount of data lost. Some existing codes successfully reduce these recovery costs but increase the storage overhead considerably. Others, which are theoretically optimal, minimize the amount of data required for recovery, but incur irregular...
[ Full version ]

Zorah Lähner and Matthias Vestner are PhD students from the group of Daniel Cremers at TU Munich. Both are working in the Analysis of Non-Rigid 3D Shapes and in particular consider the (dense) correspondence problem between instances of those. Z.L. will present (an extended version of) her CVPR 2016 paper "Efficient Globally Optimal 2D-to-3D Deformable Shape Matching" (2D-3D), M.V. will present (an extended version of) his CVPR 2017 paper "Product Manifold Filter: Non-Rigid Sha...
[ Full version ]

CS invites you to a STARTUP DAY and recruitment by the presenting firms: Augury, Axxana, CNOGA, Colu, Driveway, ENSILO, JFrog, Lightbits, SCIO, Sesame, Tabbola, Yotpo. In addition, lectures will be given by the firms representatives and entrepreneurs . The event will take place on Wednesday, March 29, 2017 between 12:30-14:30 at the CS...
[ Full version ]

This is part of our ongoing effort to develop the field of high-dimensional combinatorics. The probabilistic method and the properties of random graphs, random trees, random permutations etc. play a central role in modern combinatorics. In this talk I will discuss some of our findings concerning the higher-dimensional counterparts of these objects. My collaborators in these investigations are Zur Luria, Maya Dotan, and Michael Simkin ...
[ Full version ]

We study the problem of transferring a sample in one domain to an analog sample in another domain. Given two related domains, S and T, we would like to learn a generative function G that maps an input sample from S to the domain T, such that the output of a given function f, which accepts inputs in either domains, would remain unchanged. Other than the function f, the training data is unsupervised and consist of a set of samples from each domain. The Domain Transfer Network (DTN) ...
[ Full version ]

Virtual caches have been around for several decades. They have several advantages in performance and energy efficiency, but have not been used in ubiquitous commercial designs because of problems due to synonyms. To revisit the problem and come up with a practical design, we start with a study of the temporal behavior characteristics of synonyms in several benchmark programs. Exploiting these characteristics we propose a practical virtual cache design with dynamic synonym remappi...
[ Full version ]

You are invited to the final stage of the 2016-17 Amdocs Best Project Contest. The competing teams will present and talk about their projects. The event will take place on Tuesday, March 28, 2017, 16:30-18:30, in Auditorium 2, CS Taub Building. ...
[ Full version ]

A polyomino of area n is an edge-connected set of n cells on the square lattice. To-date, no formulae enumerating polyominoes by area (number of cells) or perimeter (number of empty cells neighboring the polyomino) are known. The area of a given polyomino will be denoted by n, and its area by p. We first prove that the maximum perimeter of a polyomino of area n is 2n+2. Then we present a few formulae enumerating polyominoes with small perimeter defect (i.e., devi...
[ Full version ]

Geometry processing deals with the design of effective discrete methods for complex problems which appear in various areas of computational science and engineering. In practice, choosing a particular discretization machinery greatly affects the formulation of the problem and the analysis and design of its computational method. Consequently, methods may differ in practical aspects such as ease of implementation and preservation of geometric features due to the choice of discretizat...
[ Full version ]

Huffman coding has a search-theoretic interpretation as the optimal strategy for the twenty questions game. In this game, Alice chooses x ∈ {1,...,n} according to a distribution µ, and Bob identifies x using yes/no questions. Bob's goal is to use the minimum number of questions in expectation. A strategy for Bob corresponds to a prefix code for {1,...,n}, and this shows that Bob's optimal strategy uses a Huffman code for µ. However, this strategy could use arbitr...
[ Full version ]

A minimum spanning tree is an essential structure for distributed algorithms, since it is a low-cost connected subgraph which provides an effcient way to communicate in a network. However, trees cannot survive even one link failure. In this talk, we study the Tree Augmentation Problem (TAP), for which the input is a graph G and a spanning tree T of G and the goal is to augment T with a minimum (or minimum weight) set of edges, such that the new graph survives any single link fail...
[ Full version ]

Learning from membership queries has flourished due to its many applications in different fields. Many of the new applications raised new models and new problems and in many of those applications the function being learned can be an arbitrary function that depends on few variables. We call this class of functions $d$-Junta, where $d$ is the number of relevant variables in the function. In some of the applications non-adaptive algorithms are most desirable, where in oth...
[ Full version ]

OpenBox is a framework that makes network functions (NFs) software-defined by decoupling their control plane from their data plane, similarly to SDN solutions that only address the network's forwarding plane. The OpenBox framework consists of a logically-centralized controller, data plane instances, and a communication protocol between them. User-defined NF applications are programmed on top of the northbound API of the controller. We present an extensible and highly modular SDN p...
[ Full version ]

Recurrent Neural Networks (RNNs) have had considerable success in classifying and predicting sequences. We demonstrate that RNNs can be effectively used in order to encode sequences and provide effective representations. The methodology we use is based on Fisher Vectors, where the RNNs are the generative probabilistic models and the partial derivatives are computed using backpropagation. State of the art results are obtained in two central but distant tasks, which both rely on seq...
[ Full version ]

A selective classifier (f,g) consists of a classification function f and a binary selection function g, which determines if the classifier abstains from prediction, or uses f to predict.The classifier is called pointwise-competitive if it classifies each point identically to the best classifier in hindsight (from the same class), whenever it does not abstain. The quality of such a classifier is quantified by its rejection mass, defined to be the probability mass of the of the poin...
[ Full version ]

This work proposes an algorithm for computing dense packings of congruent circles inside general 2D containers. Unlike the previous approaches which accept as containers, only simple, symmetric shapes such as circles, rectangles and triangles, our method works for any container with a general, freeform (spline) boundary. In contrast to most previous approaches which cast the problem into a non-convex optimization problem, our method attempts to maximize the number of ...
[ Full version ]

The 4th Technion-Intel Challenge invites you to challenge the next generation depth sensing technologies (3D cameras), deep learning techniques and IoT devices, and to the opportunity to early access to the latest Intel products and to take part in shaping the future of robotics, drones, autonomous navigation, smart homes, scene understanding, VR/AR/MR, HMI, deep learning, and more. The event will take place on Sunday, March 19, 2017, 13:00-15:00, in Room 1003, ...
[ Full version ]

The vast majority of computer users do not know how to code. Programming by examples (PBE) has flourished in recent years to address exactly this problem. PBE enables users to write their own programs by describing their intent through examples, without writing or examining a single piece of code. An inherent problem of PBE is that examples under-specify the full intent of the user. Thus, current PBE algorithms have to synthesize a program after obtaining a partial description of ...
[ Full version ]

In this research we study computation in environments with uncertainty, specifically, the distributed and streaming environments. We adapt the local-ratio technique to the distributed and streaming environments. In doing so we achieve state of the art approximation algorithms for weighted vertex cover, weighted maximum matching and weighted maximum independent set in the distributed setting. In the semi-streaming model we improve the best known approximation ratio for maximum weig...
[ Full version ]

We construct an interactive coding scheme, a notion introduced by Schulman (FOCS 1992, STOC 1993). Loosely speaking, we show how to convert any two-party interactive protocol into one that is resilient to constant-fraction of *insertion* and *deletion* errors, while preserving computational efficiency, and blowing up the communication complexity and the *round* complexity by a constant factor that approaches 0 as the error-rate approaches 0. Previous works were no...
[ Full version ]

Data matrices constructed from multimedia data are often correlated at different scales. Motivated by this observation, we consider the decomposition of a matrix into block-wise low rank components of multiple scales. We approach the problem via a convex formulation and present an iterative algorithm using block-wise SVD’s. We show that in practice, the multi-scale low rank decomposition often returns intuitive matrix decomposition. We also show results on real-world application...
[ Full version ]

A basic combinatorial interpretation of Shannon's entropy function is via the ``20 questions'' game. This cooperative game is played by two players, Alice and Bob: Alice picks a distribution $\pi$ over the numbers $\{1,\ldots,n\}$, and announces it to Bob. She then chooses a number $x$ according to $\pi$, and Bob attempts to identify $x$ using as few Yes/No queries as possible, on average. An optimal strategy for the ``20 questions'' game is given by a Huffman code for $\pi$: Bob...
[ Full version ]

The color of an object imaged in a turbid medium varies with distance and medium properties, deeming color an unstable source of information in underwater images. Assuming rough 3D scene information has become relatively easy to reconstruct, the main challenge in color reco​​very is estimating medium properties in situ, at the time of acquisition. We suggest and demonstrate a method for estimating the medium properties (both attenuation and scattering) using only images of bac...
[ Full version ]

Developing electronics beyond Moore’s Law requires revolutionary vision in novel devices, disruptive technologies, new materials and alternative computer architecture. Memristor is an emerging nanoelectronic device that use resistance instead of charge as state variable to represent digital or analog information. In this talk, I will first introduce the background and fundamental concepts about memristive devices, followed by our approaches in device engineering, fabrication/int...
[ Full version ]

Piecewise polynomial constraint systems are common in numerous problems in computational geometry, such as constraint programming, modeling, and kinematics. In this talk, we present a framework that is capable of decomposing, and efficiently solving a wide variety of complex piecewise polynomial constraint systems. The framework we present uses a constraint system decomposition algorithm to break down complex problems into smaller, simpler subprob...
[ Full version ]

Learning of layered or "deep" representations has provided significant advances in computer vision in recent years, but has traditionally been limited to fully sup​​ervised settings with very large amounts of training data. New results show that such methods can also excel when learning in sparse/weakly labeled settings across modalities and domains. I'll review state-of-the-art models for fully convolutional pixel-dense segmentation from weakly labeled input, and will discus...
[ Full version ]

Many intractable problems on graphs, can be efficiently solved for trees or forests. Tree decompositions allow taking advantage of this fact to handle general graphs by grouping nodes into bags and extracting a tree structure. The problem at hand is then solved independently for the subgraphs induced by the bags, and then the results can be efficiently combined. Tree decompositions have a plethora of applications, including join optimization in databases, constraint-satisfaction p...
[ Full version ]

A public, unpredictable source of randomness would enable many exciting applications, starting with verifiable public lotteries. It is an essential building block for many types of smart contract requiring random inputs, from online games to random audits. This talk will define this important fundamental problem and describe potentially solutions using proof-of-work based blockchains. The problem appears to require a new cryptographic primitive, the proof-of-delay: a deterministic...
[ Full version ]

Google will host a hub at CS for the Online Qualification Round of Hash Code, a team-based programming competition created by Google for university students and industry professionals. The Online Qualification Round takes place on Thursday, February 23rd at 18:30 CET and registered teams from Technion are invited to participate from our hub, which will take place at CS Taub 337. Top scoring teams from the Online Qualificatio...
[ Full version ]

Cocurrent data structures typically index data using a single primary key and provide fast access to data associated with a given key value. However, it is often required to access information via multiple primary and secondary keys, and even through additional properties that do not represent keys for the given data. We propose a lock-free and lock-based designs of a table with multiple indexing, supporting linearizable inserts, deletes and retrieve operations. We have implemente...
[ Full version ]

The cost of accessing shared objects that are stored in remote memory, while neglecting accesses to shared objects that are cached in the local memory, is evaluated by the number of remote memory references (RMRs) in an execution. two flavours of this measure- cache-coherent (CC) and distributed shared memory (DSM)-model two popular shared-memory architectures. The number of RMRs, however, does not take into account the granularity of memory accesses, namely, the fact that accesse...
[ Full version ]

Imaging through scattering media has long been a challenge, as scattering corrupts measurements in a non-invertible way. Using near-visible wavelengths to image through scattering media can realize broad applications in bio-medical and industrial imaging. It provides many advantages, such as optical contrast, non-ionizing radiation and availability of fluorescent tags. In this talk I'll discuss recent techniques that were developed to overcome and use scattering in order to recove...
[ Full version ]

Class invariants play a central role in understanding object-oriented programming. They also raise some tricky problems for the verification of OO programs, in particular "furtive access", resulting from callbacks, and "reference leak", a consequence of aliasing. I will start with a tutorial on class invariants, explaining why OO programmers should understand the concept (although today many do not even know that it exists). Then I will describe the verification issues, ...
[ Full version ]

We present a modular and demand-driven analysis of the semantic difference between program versions. Our analysis characterizes initial states for which final states in the program versions are different. It also characterizes states for which the final states are identical. Such characterizations are useful for regression verification, for revealing security vulnerabilities, and for identifying changes in the program's functionality. We are able to prove equivalence or provide...
[ Full version ]

Data-parameterized systems model finite state systems over an infinite data domain. VLTL is an extension of LTL that uses variables in order to specify properties of computations over infinite data, and as such VLTL is suitable for specifying properties of data-parameterized systems. We present alternating variable Buechi automata (AVBWs), a new model of automata over infinite alphabets, capable of modeling a significant fragment of VLTL. While alternating and non-deterministic Bu...
[ Full version ]

Roughly speaking Nano-Patterns are recurring, short snippets of code which represent a common approach for dealing with small scale implementation issues. Examples include, e.g., ‟\emph{set and return old value}”, ‟\emph{defaults to}”, and the sequence of instructions required to implement the logical condition~$∃x∈S∙p(x)$ where~$S$ is a collection and~$p(·)$ is a logical predicate. This paper presents the notion of Nano-Patterns, offers criteria for evaluating the...
[ Full version ]

Geographic search, where the user provides keywords and receives relevant locations depicted on a map, is a popular web application. Typically, such a search is based on static geographic data. However, the abundant geotagged posts in microblogs such as Twitter and in social networks like Instagram provides contemporary information that can be used to support geosocial searches. Geographic searches based on user activities in social media. Such searches can point out where peop...
[ Full version ]

Images taken in scattering media, such as haze, fog, and underwater, often look faded and lack contrast. We detect changes in pixels' distribution in RGB space due to the scattering medium: instead of tight clusters the pixels form lines, which we term Haze-Lines. In this talk the model will be introduced, and two applications will be discussed. First, estimating the air-light: the color of a pixel with no object in the line of sight. Second restoring the colors of underwater imag...
[ Full version ]

The direct-sum question in two-party communication complexity is the following; Alice receives $(x_1,\dots,x_\ell)$ and Bob receives $ (y_1,\dots,y_\ell) $, where each $x_i$ and $y_i$ are $n$-bit strings. Together, they wish to compute $f(x_i,y_i)$ for every $ i \in \{1,\dots,\ell\}$, where $f$ is some predetermined function. A \emph{saving} is said to occur if Alice and Bob can utilize the fact that they are given the $\ell$ instances simultaneously in order to compute the outpu...
[ Full version ]

Using image context is an effective approach for improving object detection. Previously proposed methods used contextual cues that rely on semantic or spatial information. In this work, we explore a different kind of contextual information: inner-scene similarity. We present the CISS (Context by Inner Scene Similarity) algorithm, which is based on the observation that two visually similar sub-image patches are likely to share semantic identities, especially when both appear in the...
[ Full version ]

Algebraic constraints arise in various applications, across domains in science and engineering. Polynomial and piece-wise polynomial (B-Spline) constraints are an important class, frequently arising in geometric modeling, computer graphics and computer aided design, due to the useful NURBs representation of the involved geometries. Subdivision based solvers use properties of the NURBs representation, enabling, under proper assumptions, to solve non-linear, multi-variate algebraic ...
[ Full version ]

Fast and robust three-dimensional reconstruction of facial geometric structure from a single image is a challenging task with numerous applications in computer vision and graphics. We propose to leverage the power of convolutional neural networks (CNNs) to produce highly detailed face reconstruction directly from a single image. For this purpose, we introduce an end-to-end CNN framework which constructs the shape in a coarse-to-fine fashion. The proposed architecture is composed o...
[ Full version ]

A polyomino of size n consists of n squares joined along their edges. A popular example is the computer game Tetris, which features polyominoes of size 4. A d-dimensional polycube of size n is a connected set of n d-dimentional cubes, where connectivity is through (d−1) dimensional faces. Fixed polycubes are usually considered identical if one can be translated into the other. In this research we focus on basic questions along the lines of: how many fixed d-dimentional polycubes...
[ Full version ]

Yahoo will hold the annual event at Technion CS on Wednesday, January 25, 2017, 14:30-16:30, in room 337 (3rd floor) of the CS Taub Building The program includes introduction words by VP Yoelle Maarek, followed by Research Director Dan Pelleg's lecture on "Automatic Trivia Fact Extraction from Wikipedia", after which you will be invited to Research Round Tables and refreshment. Full program and more details in the attached ad. You...
[ Full version ]

Obstruction-free consensus, ensuring that a process running solo will eventually terminate, is at the core of practical ways to solve consensus, e.g., by using randomization or failure detectors. An obstruction-free consensus algorithm may not terminate in many executions, but it must terminate whenever a process runs solo. Such an algorithm can be evaluated by its solo step complexity, which bounds the worst case number of steps taken by a process running alone, from any configur...
[ Full version ]

Modern computerized systems are complex designs that include hardware and software components. Designing and implementing such systems requires extensive engineering. Yet, unlike other domains of engineering, software and hardware engineers often lack the mathematical tools to help them specify the requirements and verify that the implementation conforms to the specification. Formal Methods aim at bridging this gap. In particular, Formal Verification (FV) techniques supply the too...
[ Full version ]

On Tuesday, January 24, 2017, the Systems and Software Development Laboratory (SSDL) will hold a project Fair on IoT and Android, presenting the newest and most inspiring projects presented by the developing teams. You are all invited! Following are the presenting projects: ANDROID BicyCare...
[ Full version ]

I will present out approach to human pose estimation, where each location in the image votes for the position of each keypoint using a convolutional neural net. The voting scheme allows us to utilize information from the whole image, rather than rely on a sparse set of keypoint locations. Using dense, multi-target votes, not only produces good keypoint predictions, but also enables us to compute image-dependent joint keypoint probabilities by looking at consensus voting. ...
[ Full version ]

Algorithms and the Internet are revolutionizing "markets" - the mechanisms through which resources are allocated among players under optimization criteria. While resource allocation is a long-standing theme in the study of classic algorithms like matching and routing, the need to interact with self-interested players, and the uncertain inputs they provide, break traditional algorithms and raise fundamental new challenges. Applications include the allocation of cloud computin...
[ Full version ]

Batch codes, first introduced by Ishai, Kushilevitz, Ostrovsky, and Sahai, mimic a distributed storage of a set of n data items on m servers, in such a way that any batch of k data items can be retrieved by reading at most some t symbols from each server. Combinatorial batch codes, are replication-based batch codes in which each server stores a subset of the data items. In this talk, we propose a generalization of combinatorial batch codes, called multis...
[ Full version ]

Modern computerized systems are complex designs that include hardware and software components. Designing and implementing such systems requires extensive engineering. Yet, unlike other domains of engineering, software and hardware engineers often lack the mathematical tools to help them specify the requirements and verify that the implementation conforms to the specification. Formal Methods aim at bridging this gap. In particular, Formal Verification (FV) techniques ...
[ Full version ]

My work focuses on ways of replicating data in demanding settings, most recently the cloud. The cloud is a setting where copying information and replicating data or computation is common, yet it remains difficult to actually create new applications that leverage replication. Moreover, the existing libraries are very slow. I’ll present Derecho, a blazingly fast C++ library for creating scalable data replication solutions. Derecho has a remarkably simple API and ...
[ Full version ]

For every given real value epsilon>0, a Fully Polynomial Time Approximation Scheme (FPTAS) computes in polynomial time (in both the input size and 1/epsilon) a feasible solution that is close to the optimal solution within ratio epsilon. As epsilon can be chosen arbitrary small, and the running time is polynomial, FPTASs are considered as the “Holy grail” of approximation algorithms, but most optimization problems, e.g., strongly NP-hard problems, cannot admit an FPTAS unless ...
[ Full version ]

Crowd Mining is concerned with identifying significant patterns in the knowledge of the crowd, capturing, e.g., habits and preferences, by posing internet users with targeted questions. To account for jointly processing the crowd answers and available knowledge bases, and for user interaction and optimization issues, crowd mining frameworks must employ complex reasoning, automatic crowd task generation and crowd member selection. In this talk I will present the unique challenges i...
[ Full version ]

...
[ Full version ]

Efficient proof verification is at the heart of the study of computation. Seminal results such as the IP=SPACE Theorem [LFKN92,Shamir92] and the PCP theorem [AS92,ALMSS92] show that even highly complicated statements can be verified extremely efficiently. We study the complexity of proving statements using interactive protocols. Specifically, what statements can be proved by a polynomial-time prover to a super-efficient verifier. Our main results show that these pro...
[ Full version ]

Software metrics are used by software engineers to help gauge the health of their projects. Researchers hope to correlate easy to measure properties, such as lines of code, cyclomatic complexity, and the number of operators and operations, with external, harder to measure properties such as maintainability and proneness to bugs. Over the years, hundreds of metrics have been proposed; unsurprisngly, most metrics are correlated with the size of the code module. However, it isn't cle...
[ Full version ]

We give a new deterministic algorithm that non-adaptively learn a hidden hypergraph from edge-detecting query. This algorithm use a very interesting algebraic families: Perfect Hash, Universal Set and Cover Free family. All previous non-adaptive algorithms either run in exponential time or have non optimal query complexity. We give the first polynomial time non-adaptive learning algorithm for learning hypergraph that asks an almost optimal number of queries. ...
[ Full version ]

Classical AI planning is concerned with the following problem: Given a deterministic system, an initial system state, and a goal condition, find a sequence of actions that transforms the system from its initial state to a state that satisfies the goal condition. It was originally conceived in order to make robots autonomous, and has numerous applications. A simple and natural extension of classical planning is one where there are multiple agents, each with its own set of act...
[ Full version ]

Programming languages have many similarities, and so, when writing a source-to-source transformation on one language, it would be nice to reuse code from a similar transformation for a different language. This is a fundamentally difficult problem, and previous attempts have either resorted to reimplementing the same transformation for many languages, or at best reducing multiple languages to a common intermediate representations, which necessarily destroys information and produces...
[ Full version ]

The class P attempts to capture the efficiently solvable computational tasks. It is full of practically relevant problems, with varied and fascinating combinatorial structure. In this talk, I will give an overview of a rapidly growing body of work that seeks a better understanding of the structure within P. Inspired by NP-hardness, the main tool in this approach are combinatorial reductions. Combining these reductions with a small set of plausible conjectures, we obtain t...
[ Full version ]

I will describe two branches of my work related to algorithms for distributed networks. The main focus will be devoted for Fault-Tolerant (FT) Network Structures. The undisrupted operation of structures and services is a crucial requirement in modern day communication networks. As the vertices and edges of the network may occasionally fail or malfunction, it is desirable to make those structures robust against failures. FT Network Structures are low cost highly resili...
[ Full version ]

Computing the epipolar geometry between cameras with very different viewpoints is often problematic as matching points are hard to find. In these cases, it has been proposed to use information from dynamic objects in the scene for suggesting point and line correspondences. We introduce an approach that improves by two orders of magnitude the performance over state-of-the-art methods, by significantly reducing the number of outliers in the putative matches. Our approach is based ...
[ Full version ]

Motivation: Isotope tracing coupled with Metabolic Flux Analysis (MFA) is a commonly used approach for quantifying cellular metabolic fluxes. Isotope tracing involves feeding cells with isotopic labeled nutrients and tracking the labeling of metabolites via mass spectrometry and NMR. MFA computationally analyzes these isotopic measurements to infer flux. A major limitation of MFA is its strict reliance on computationally hard non-convex optimizations, requiring heuristic solv...
[ Full version ]

In this talk, we show an explicit construction of extractors for two independent sources of near-logaritmic min-entropy. Previous constructions required either polylog(n) min-entropy or more than two sources. The result extends the breakthrough result of Chattopadhyay and Zuckerman and also uses non-malleable extractors. The main new ingredient is a somewhere-random condenser with a small entropy gap, used as a sampler. Our...
[ Full version ]

...
[ Full version ]

The proliferation of data collection in the health, commercial, and economic spheres, brings with it opportunities for extracting new knowledge with concrete policy implications. Examples include individualizing medical practices based on electronic healthcare records, and understanding the implications of job training programs on employment and income. The scientific challenge lies in the fact that standard prediction models such as supervised machine learning are ...
[ Full version ]

The past five years have seen a dramatic increase in the performance of recognition systems due to the introduction of deep architectures for feature learning and classification. However, the mathematical reasons for this success remain elusive. In this talk we will briefly survey some existing theory of deep learning. In particular, we will focus on data structure based theory and discuss two recent developments. The first work studies the generalization error of deep...
[ Full version ]

The proliferation of data collection in the health, commercial, and economic spheres, brings with it opportunities for extracting new knowledge with concrete policy implications. Examples include individualizing medical practices based on electronic healthcare records, and understanding the implications of job training programs on employment and income. The scientific challenge lies in the fact that standard prediction models such as supervised machine learning are often no...
[ Full version ]