The Taub Faculty of Computer Science Events and Talks

Deep learning has become pervasive in everyday life, powering language applications like Apple's Siri, Amazon's Alexa, and Google Translate. The inherent limitation of these deep learning systems, however, is that they often function as a ''black box'', preventing researchers and users from discerning the roles of different components and what they learn during the training process. In this talk, I will describe my research on interpreting deep learning models for language along three lines. First, ...
[Full version]

The 2018 Workshop in Cryptology, organized by the Technion Hiroshi Fujiwara Cyber Security Research Center, will be held on Monday, December 31 2018, between 9:00-16:20, in the Auditorium of the Water Institute, Technion. Most talks will be given in Hebrew but some in English and during the break a poster presentation will be held: Graduate students and all who are interested to present their research are welcome to book their post as soon as possible ...
[Full version]

For a set M = {−µ, −µ + 1, . . . , λ} \ {0} with non-negative integers λ, µ Small covering sets play an important role in codes correcting limited-magnitude errors. We use some number theoretic methods to give an explicit construction of a (λ, µ; q)-covering set S which is of the size q1+o(1) max{λ, µ}−1/2 for almost all integers q ? 1 and of optimal order of magnitude (that is up ...
[Full version]

In formal verification, one uses mathematical tools in order to prove that a system satisfies a given specification. A limitation of traditional formal-verification algorithms and tools concerns the certification of positive results: while a verifier may answer that a system satisfies its specification, a designer often needs some palpable evidence, or certificate, of correctness. I will discuss the notion of certificates in several applications of formal verification, and present two works addressing the above limitation ...
[Full version]

Edit distance is a measure of similarity of two strings based on the minimum number of character insertions, deletions, and substitutions required to transform one string into the other. The edit distance can be computed exactly using a dynamic programming algorithm that runs in quadratic time. Andoni, Krauthgamer and Onak (2010) gave a nearly linear time algorithm that approximates edit distance within approximation factor poly(log n). In this talk, I'll present a recent result in ...
[Full version]

Sediment resuspension is the transport of previously settled particles from the seafloor back into the overlying water. Measuring these abrupt and spatially varying events is challenging. Existing in-situ approaches are very localized. We presented a novel underwater wide field imaging system designed to (a) observe the seafloor and the water medium above it from a distance, (b) sense sediment resuspension events, and (c) algorithmically quantify the resuspension. Our technology quantifies the amount of material lifted, ...
[Full version]

Distributed training can significantly reduce the training time of neural networks. Despite its potential, however, distributed training has not been widely adopted due to the difficulty of scaling the training process. Existing methods suffer from slow convergence and low final accuracy when scaling to large clusters, and often require substantial re-tuning of hyper-parameters. We propose DANA, a novel approach that scales to large clusters while maintaining similar final accuracy and convergence speed to that of ...
[Full version]

Constrained codes have been studied actively for a long time with numerous applications in data recording and data communication systems. Recently, various types of one-dimensional and two-dimensional constrained codes have attracted significant attention owing to some applications for flash memories. ICI-free codes (Amit Berman and Yitzhak Birk - 2010), balanced q-ary ICI-free codes (Minghai Qin, Eitan Yaakobi and Paul Siegel - 2014), and weakly constrained codes(Sarit Buzaglo and Paul Siegel - 2017) are such codes. ...
[Full version]

Despite their superb empirical performance, deep learning models for natural language processing (NLP) are often considered black boxes, as relatively little is known as to what accounts for their success. This lack of understanding turns model development into a slow and expensive trial-and-error process, which limits many researchers from developing state-of-the-art models. Customers of deep learning also suffer from this lack of understanding, because they are using tools that they cannot interpret. In this talk ...
[Full version]

In the classical Node-Disjoint Paths (NDP) problem, we are given an n-vertex graph G, and a collection of pairs of its vertices, called demand pairs. The goal is to route as many of the demand pairs as possible, where to route a pair we need to select a path connecting it, so that all selected paths are disjoint in their vertices. The best current algorithm for NDP achieves an $O(\sqrt{n})$-approximation, while, until recently, the best ...
[Full version]

Optimal transport (OT) has become a fundamental mathematical tool at the interface between calculus of variations, partial differential equations and probability. It took however much more time for this notion to become mainstream in numerical applications. This situation is in large part due to the high computational cost of the underlying optimization problems. There is a recent wave of activity on the use of OT-related methods in fields as diverse as computer vision, computer graphics, ...
[Full version]

A primitive k-batch code encodes a string x of length n into string y of length N, such that each multiset of k symbols from x has k mutually disjoint recovering sets from y. The definition of such codes is motivated by applications to load balancing in distributed storage and private information retrieval. We develop new constructions of linear primitive batch codes based on finite geometries. In some parameter regimes, our codes have lower redundancy ...
[Full version]

As software has grown increasingly critical to our society's infrastructure, mechanically-verified software has grown increasingly important, feasible, and prevalent. Proof assistants have seen tremendous growth in recent years because of their success in the mechanical verification of high-value applications in many areas, including cyber security, cyber-physical systems, operating systems, compilers, and microkernels. These proof assistants are built on top of constructive type theory whose computational interpretation is given by the proofs-as-programs paradigm, which establishes a ...
[Full version]

The tremendous success of the Machine Learning paradigm heavily relies on the development of powerful optimization methods. The canonical algorithm for training learning models is SGD (Stochastic Gradient Descent), yet this method has its limitations. It is often unable to exploit useful statistical/geometric structure, it might degrade upon encountering prevalent non-convex phenomena, and it is hard to parallelize. In this talk I will discuss an ongoing line of research where we develop alternative methods that ...
[Full version]

Distributed Denial of Service (DDoS) attacks are doubling in both number and volume on a yearly basis. These pose a critical threat to financial institutions and cloud providers that struggle to keep their services available and secure. To mitigate the attacks, operators rely on middleboxes that analyze the traffic and identify malicious flows and subnets. A key technique used for this identification is the Hierarchical Heavy Hitters (HHH) measurement, that singles out networks which send ...
[Full version]

As algorithmic prediction systems have become more widespread, so too have concerns that these systems may be discriminatory against groups of people protected by laws and ethics. We present a recent line of work that takes a complexity theoretic perspective towards combating discrimination in prediction systems. We'll focus on fair classification within the versatile framework of Dwork et al. [ITCS'12], which assumes the existence of a metric that measures similarity between pairs of individuals. Unlike ...
[Full version]

Multi-access Edge Computing (MEC) is a new network architecture that allows applications and network services to be executed at the edge of the network. This is done by running these services on commodity servers that are placed in close proximity to the network edge and to the cellular base stations in wireless networks. This architecture provides high bandwidth and low latency for network functions and other applications. However, the availability of the resources at the ...
[Full version]

We all capture the world around us through digital data such as images, videos and sound. However, in many cases, we are interested in certain properties of the data that are either not available or difficult to perceive directly from the input signal. My goal is to “Re-render Reality”, i.e., develop algorithms that analyze digital signals and then create a new version of it that allows us to see and hear better. In this talk, ...
[Full version]

Sparse coding refers to the pursuit of the sparsest representation of a signal in a typically overcomplete dictionary. From a Bayesian perspective, sparse coding provides a Maximum a Posteriori (MAP) estimate of the unknown vector under a sparse prior. Various nonlinear algorithms are available to approximate the solution of such problems. In this work, we suggest enhancing the performance of sparse coding algorithms by a deliberate and controlled contamination of the input with random noise, ...
[Full version]

List-decoding is an important primitive in the theory of error correcting codes, and it has long been a goal to obtain explicit constructions of capacity-achieving, efficiently list-decodable codes. Folded Reed-Solomon Codes (Guruswami-Rudra 2008) and Multiplicity codes (Guruswami-Wang 2011, Kopparty 2012) are two such constructions. However, previous analysis of these codes could not guarantee optimal parameters. In particular, the “list-size” of these codes was only shown to be polynomial, while ideally it would be constant. Over ...
[Full version]

Today, extrusion is a one of the most important techniques for manufacturing continuous polymer profiles. Despite their widespread use, extruded products initially often suffer from both poor shape quality and material inhomogeneities due to inadequate process design. In an industrial environment, the elements of the extrusion line are therefore subject to extensive running-in trials. During these trials, both geometry and process conditions are iteratively adapted until the product attains the required quality. It is the ...
[Full version]

We introduce load-balanced fractional repetition (LBFR) codes, which are a strengthening of fractional repetition (FR) codes. LBFR codes have the additional property that multiple node failures can be sequentially repaired by downloading no more than one block from any other node. This allows for better use of the network, and can additionally reduce the number of disk reads necessary to repair multiple nodes. We characterize LBFR codes in terms of their adjacency graphs, and use ...
[Full version]

We explore the power of interactive proofs with a distributed verifier. In this setting, the verifier consists of $n$ nodes and a graph $G$ that defines their communication pattern. The prover is a single entity that communicates with all nodes by short message. The goal is to verify that the graph $G$ belongs to some language in a small number of rounds, and with small communication bound, \ie the proof size. This interactive model was ...
[Full version]

Cryptographic hash functions are the basis of many important and far reaching results in cryptography, complexity theory, and beyond. In particular, hash functions are the primary building block of fundamental applications like digital signatures and verifiable computation, and they are the tool underlying the proofs-of-work which drive blockchains. Because of the central role of cryptographic hash functions in both theory and practice, it is crucial to understand their security guarantees, toward basing applications on the ...
[Full version]

In this talk, we will present error-correcting codes for the storage of data in synthetic deoxyribonucleic acid (DNA). We investigate a storage model where a data set is represented by an unordered set of M sequences, where each sequence is a vector of length L over Z_q. Errors within that model are losses of whole sequences and point errors inside the sequences, such as insertions, deletions and substitutions. We derive lower bounds on the minimum ...
[Full version]

We propose a novel discrete scheme for simulating viscous thin films at real-time frame rates. Our scheme is based on a new formulation of the gradient flow approach, that leads to a discretization based on local stencils that are easily computable on the GPU. Our approach has physical fidelity, as the total mass is guaranteed to be preserved, an appropriate discrete energy is controlled, and the film height is guaranteed to be non-negative at all ...
[Full version]

Quantum computers may be able to solve certain problems that are so complicated that even the fastest supercomputer would take millions of years to provide an answer. Entanglement and superposition are quantum phenomena which can be tamed in order to build such a machine. Optimising financial transactions, machine learning, creating new medicines, understanding protein folding and breaking codes are just some of the problems where the existence of a quantum computer could change everything. The ...
[Full version]

Consider a directed analogue of the random graph process on $n$ vertices, whereby the $n^2-n$ directed edges are ordered uniformly at random and revealed one at a time, giving a nested sequence of directed graphs $D_0,D_1,\dots,D_m, \dots$. In this setting, one may ask about events that hold with probability $1-o(1)$ (whp) as $n$ tends to infinity. In particular, for a fixed $q=O(1)$, we wish to study the hitting time for the emergence of $q$ edge-disjoint ...
[Full version]

The steady rise of intelligence and autonomy over a scale of distributed, connected and smart components is heralding the era of Internet-of-Intelligence. Without adequate safeguards in place, such components can lead to terrible consequences. In this talk, I will discuss three aspects - namely, vulnerability, security and privacy of such edge–computing scenarios. The vulnerability aspect will show how a simple setup can derail sophisticated unmanned aerial vehicles. The security angle will be discussed through a ...
[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 ...
[Full version]

Going from driving assistance to autonomous driving, requires a deeper understanding of the surroundings of the vehicle. Driving assistance systems provide technological solutions that help with the driving process, while these systems usually provide multiple features that enhance driving safety (such as automate lighting, adaptive cruise control, collision avoidance, etc.), they are still very simplistic with respect to their level of scene understanding. Autonomous driving, raises much more complex problems: Is the detected person a ...
[Full version]

Program synthesis, the task of automatically finding a program that satisfies a given user intent specification, has been successfully applied to aid commercial data wrangling, software engineering, and question answering. While data-driven (deep learning) and symbolic (formal logic) techniques are both commonly used for program synthesis, both of them have their strengths and weaknesses. Symbolic techniques guarantee correctness of the generated program with respect to the specification, but are difficult to design, develop, and tune. ...
[Full version]

One of the most promising error correcting schemes is the family of Spatially-Coupled (SC) LDPC codes. SC-LDPC codes achieve capacity universally on memoryless binary symmetric channels, and their special structure can be exploited to implement low-latency high-throughput decoders. In this talk, a new type of SC-DPC codes motivated by practical storage applications is presented. These new codes can be decoded locally at the level of sub-blocks that are much smaller than the full code block, ...
[Full version]

We show that bounded independent distributions under perturbation of noise fool product tests, which are products of bounded functions defined on disjoint inputs. These results have found applications in coding theory and pseudorandomness. In this talk, I will talk about communication and space lower bounds for decoding linear codes, and how to construct pseudorandom generators for product tests and space-bounded computation. Based on joint works with Elad Haramaty and Emanuele Viola.
[Full version]

There is a widening gap in the datacenter between data growth and stagnating CPU performance. This gap limits our ability to solve more complex problems and prompts us to revisit both the architecture of servers and the ways we manage and process data. In my work, I aim to narrow this gap using specialization and HW/SW co-design. As a specific example, I will talk about building energy-efficient distributed storage for large-scale data processing applications. Most ...
[Full version]

The emergence of very effective deep learning techniques in recent years has affected almost all areas of research remotely related to AI, and computer vision in particular has been changed irreversibly. In this talk I will focus on visual object recognitions. The incredible recent progress in this area, and the availability of very effective public domain tools for object recognition in images, allows us to reopen old questions and approach them from new directions with ...
[Full version]

Trusted execution environments, and particularly the Software Guard eXtensions (SGX) included in recent Intel x86 processors, gained significant traction in recent years. A long track of research papers, and increasingly also real-world industry applications, take advantage of the strong hardware-enforced confidentiality and integrity guarantees provided by Intel SGX. Ultimately, enclaved execution holds the compelling potential of securely offloading sensitive computations to untrusted remote platforms. In this talk, I will present Foreshadow, a practical software-only microarchitectural ...
[Full version]

Codes with locality and/or availability have been extensively studied in recent years, including locally repairable codes (LRC), PIR codes, batch codes, etc. Usually in such a code of dimension s, we focus on the recovering sets only for the s information symbols. We propose a natural generalization of PIR and batch codes, named functional PIR codes and functional batch codes, by analyzing the recovering sets for arbitrary vectors of length s. In this talk we ...
[Full version]

Concurrent data structures are widely used in modern multi-core architectures, providing atomicity (linearizability) for each concurrent operation. However, it is often desirable to execute several operations on multiple data structures atomically. We present a design of a transactional framework supporting linearizable transactions of multiple operations on multiple data structures in a lock-free manner. Our design uses a helping mechanism to obtain lock-freedom, and an advanced lock-free contention management mechanism to mitigate the effects of aborting ...
[Full version]

or a vector space F^n over a field F, an (η, ß)-dimension expander of degree d is a collection of d linear maps Γ_j : F^n \to F^n such that for every subspace U of F^n of dimension at most ηn, the image of U under all the maps, ∑_{j=1}^d Γ_j(U), has dimension at least ßdim(U). Over a finite field, a random collection of d=O(1) maps Γ_j over excellent “lossless” expansion with high probability: ß ...
[Full version]

Global localization for robotic vehicles is an essential backbone for robust autonomous navigation. While GPS systems offer effective solutions in outdoor environments, they are inadequate for providing positioning in indoor settings. In order to solve this problem, we investigate a special vision-based approach. Vision systems are particularly advantageous due to the amount of information they supply about the scene. To offer a system which has low cost and fast setup, we utilize a monocular camera. ...
[Full version]

Machine learning (ML) has deeply impacted many areas of computer science, including computer vision, natural language processing, computational biology, and beyond. Yet, computer networking has largely withstood the ML tide until recently. Recent advances suggest that this might be changing. We ask whether/when traditional network protocol design, which traditionally relies on the application of algorithmic insights by human experts, can be replaced by a data-driven, ML-guided approach. We will investigate this question in the context ...
[Full version]

Deep learning has become an extremely effective tool for image classification and image restoration problems. Here, we apply deep learning to microscopy, and demonstrate how neural networks can exploit the chromatic dependence of the point-spread function to classify the colors of single emitters imaged on a grayscale camera. While existing single-molecule methods for spectral classification require additional optical elements in the emission path, e.g. spectral filters, prisms, or phase masks, our neural net correctly identifies ...
[Full version]

At its core, coding theory studies how many elements of a (finite) vector space one can pack subject to the constraint that the elements are well-spread. Typically, the notion of closeness is that of Hamming distance, that is, the distance between two vectors is the number of coordinates on which they differ. In a rank-metric code, codewords are matrices over a finite field and the distance between codewords is the rank of their difference. A ...
[Full version]

In this talk I will discuss a new tangent estimator for 3D digital curves. The proposed estimator is based on 3D digital line recognition, and it is an extension of a similar 2D tangent estimator proposed for tangent estimating along digital contours. The main advantages of this new tangent estimator are its speed and its asymptotic convergence – the estimated tangents converge to the ground truth as the resolution increases.
[Full version]

We consider the following simple random experiment to determine the impact of concurrency on the performance of binary search trees: n randomly permuted keys arrive one at a time. When a new key arrives, it is first placed into a buffer of size c. Whenever the buffer is full, or when all keys have arrived, an adversary chooses one key from the buffer and inserts it into the binary search tree. The ability of the ...
[Full version]

Data intensive tasks have been ubiquitous ever since the data science revolution. The immensity of contemporary datasets no longer allows computations to be done on a single machine, and distributed computations are inevitable. Since most users cannot afford to maintain a network of commodity servers, burdensome computations are often outsourced to third party cloud services. However, this approach opens a Pandora's box of potential woes, such as malicious intervention in computations, privacy infringement, and workload ...
[Full version]

Large points cloud X in $R^{n\times D}$ are often assumed to be sampled from a k-dimensional manifold where $k 1$). However, there is no evidence that this technique extends to other manifolds. This work aims to verify how the multi-scale singular value analysis of a manifold can extend to any manifold. In this work, we focus our effort on signal manifolds of super-imposed pulses (SIPS), due to their generic nature and widespread use in signal ...
[Full version]

Pseudorandom functions (PRFs) are one of the fundamental building blocks in cryptography. Traditionally, there have been two main approaches for PRF design: the "practitioner's approach" of building concretely-efficient constructions based on known heuristics and prior experience, and the "theoretician's approach" of proposing constructions and reducing their security to a previously-studied hardness assumption. While both approaches have their merits, the resulting PRF candidates vary greatly in terms of concrete efficiency and design complexity. In this work, ...
[Full version]

Medical ultrasound (US) is a widespread imaging modality owing its popularity to cost efficiency, portability, speed, and lack of harmful ionizing radiation. At the same time, there are trade-offs among different US imaging qualities such as frame-rate, resolution, signal-to-noise-ratio and contrast. So far, these trade-offs have been compensated by mostly traditional model-based signal-processing methods. In the wake of the recent revival of artificial neural networks (NN), or more specifically, deep convolutional neural networks (CNN) for ...
[Full version]

Scaling of computing performance enables new applications and greater value from computing. With the end of Moore’s Law and Dennard Scaling, continued performance scaling will be due primarily to specialization. Graphics processing units are an ideal platform on which to build domain-specific accelerators. They provide very efficient, high performance communication and memory subsystems - which are needed by all domains. Specialization is provided via “cores”, such as tensor cores or ray-tracing cores that accelerate specific ...
[Full version]

The Convolutional Sparse Coding (CSC) model has recently gained considerable traction in the signal and image processing communities. By providing a global, yet tractable, model that operates on the whole image, the CSC was shown to overcome several limitations of the patch-based sparse model while achieving superior performance in various applications. Contemporary methods for learning the CSC dictionary often rely on the Alternating Direction Method of Multipliers (ADMM) in the Fourier domain for the computational ...
[Full version]

In many Natural Language Processing classification tasks, the label space consists of the entire vocabulary, and therefore might have hundreds of thousands of labels. Important tasks such as language modeling, machine translation and dialog systems all have vocabulary label sets. Due to Zipf's law, a large number of words in the vocabulary will have only a few appearances in the corpus, hindering the ability to learn proper representations for these words. This work utilizes a ...
[Full version]

We consider the pose-estimation problem of aligning (rotating and translating) a set of n points to a corresponding set of n lines, both on the plane. The goal is to minimize the sum of distances between the matched point-line pairs over all possible translations, rotations, and matching functions. These natural problems occur e.g. in localization of GPS data (aligning points to a map), 2D images (object to RGB pixels) or stars (to sky patterns, e.g. ...
[Full version]

The Hiroshi Fujiwara Cyber Security Research Center will hold the 7th Summer School on Cyber and Computer Security: "Trusted Execution and Hardware Side Channels". The event will take place on Tuesday-Thursday, October 2nd-4th, 2018 at the Technion, Haifa. Chairs: Mark Silberstein, Technion Yossi Oren, Ben-Gurion University Speakers: Ittai Anati, Intel Israel Herbert Bos, VU Amsterdam Christof Fetzer, TU Dresden Daniel Genkin, University of Michigan Taesoo Kim, Georgia Tech Eyal Ronen, Weismann Institute of Science Avishai ...
[Full version]

This work investigates algorithms and data structures for volumetric representation (V-reps) of 3D objects, representing the interior of the object in addition to its boundaries, extending the contemporary Boundary representation (B-rep) common scheme. In recent years, there is a growing and emerging need for a volumetric representation of 3D objects. Specifically, with the development of Iso-geometric Analysis (IGA) and advanced manufacturing technologies employing heterogeneous materials, such as 3D-printing and additive manufacturing (AM) of functionally graded ...
[Full version]

A d-dimensional polycube is a facet-connected set of cells (cubes) on the d-dimensional cubical lattice. Let Ad(n) denote the number of d-dimensional polycubes (distinct up to translations) with n cubes, and λd denote the limit of the ratio Ad(n+1)/Ad(n) as n approaches infinity. The exact value of λd is still unknown rigorously for d ≥ 2; the asymptotics of λd, as d approaches infinity, also remained elusive as of today. In this talk, I will ...
[Full version]

Using a mold or die, primary shaping manufacturing processes form material from an initially unshaped state (usually melt) into a desired shape. Examples of such a process are extrusion or high-pressure die casting — processes that are responsible for many products of our everyday life, from pipes to yoghurt cups. From the design perspective, what these processes have in common is that the exact design of the mold cannot be determined directly and intuitively from ...
[Full version]

Cellular metabolic demands change throughout the cell cycle. Nevertheless, a characterization of how metabolic fluxes adapt to the changing demands throughout the cell cycle is lacking. The rate of metabolic reactions and pathways in living cells, also referred to as metabolic flux, is not a directly measurable quantity. The most direct approach for quantifying intracellular metabolic flux is isotope tracing coupled with computational metabolic flux analysis. This has become a central technique in studies of ...
[Full version]

As the cyber security threat continues to grow, we need to better defend our systems. One of the ways to improve the security of a computer system is to isolate its critical assets from the main execution environment. These critical assets are handled in a Trusted Execution Environment, which can protect them even when the main execution environment is compromised. One of the leading technologies enabling trusted execution environment is ARM TrustZone. In this talk ...
[Full version]

Recent advances in deep learning and convolutional neural networks (CNNs) have had a profound impact on almost every computer vision task. However, generic (non-semantic) image segmentation is a notable exception despite it being one of the most fundamental and widely studied tasks in this field. In this talk, we revisit the generic segmentation task and propose Deep Generic Segmentation (DGS) -- a new deep learning approach combined with conditional random fields (CRFs). Our method differs ...
[Full version]

Moving from the concept of discrete sequences of 2D slides towards smooth 3D multimodal hierarchical presentations promises many improvements in quality and effectiveness of presentations. On the other hand, such a move poses many difficulties, one of which is how to arrange content in a 3D space. This task becomes further complicated when the story-graph of the presentation is evolving and is more complex than a single linear story-path. In this work, we propose a ...
[Full version]

Bluetooth is a widely deployed platform for wireless communications between mobile devices. It uses authenticated Elliptic Curve Diffie-Hellman for its key exchange. We show that the authentication provided by the Bluetooth pairing protocols is insufficient and does not provide the promised MitM protection. We present a new variant of an Invalid Curve Attack that preserves the x-coordinate of the public keys. The attack compromises the encryption keys of all of the current Bluetooth authenticated pairing ...
[Full version]

Designing a new drug is an expensive and lengthy process. The first stage is drug discovery, in which potential drugs are identified before selecting a candidate drug to progress to clinical trials. As the space of potential molecules is very large (10^23-10^60), a common technique during drug discovery is to start from a molecule which already has some of the desired properties. An interdisciplinary team of scientists generates hypothesis about the required changes to the ...
[Full version]

Text processing on the web is challenging due to the use of informal and ungrammatical language. Yet, this is a very prominent domain for Natural Language Processing, due to the growing volume of textual information that is communicated through Internet pages and social media platforms. In this thesis we describe a particular form of text processing on the web - syntactic parsing of web queries. We present two novel contributions: (1) We extend the transition ...
[Full version]

Efficient and sustainable conversion of biomass into valuable products is a major challenge for bioengineering. The composition of the feedstock biomass and the ability of microorganisms to efficiently ferment it are two most critical factors influencing the process efficiency. Intelligent design that addresses both these factors can greatly benefit from organism metabolic models and from using them in simulations and in computer-assisted optimization of the fermentation processes. In this talk we will cover several aspects ...
[Full version]

In Attribute-Based Signatures an authority can generate multiple signing keys, where each key is associated with a constraint f. A key respective to f can sign a message x only if f(x)=0. The security requirements are unforgeability and key privacy (signatures should not expose the specific signing key used). In Homomorphic Signatures, given a signature for a data-set x, one can evaluate a signature for the pair (f(x),f), for functions f. In context-hiding HS, evaluated ...
[Full version]

A coreset (or core-set) of a dataset is its semantic compression with respect to a set of classifiers, such that learning the (small) coreset provably yields an approximate classifier of the original (full) dataset. Composable coresets also allow handling streaming and distributed data, e.g. using a cloud or swarm of drones. In this talk I will forge a link between coresets for fundamentals problems in machine (active and deep) learning -- to problems in real-time ...
[Full version]

CS Labs invited you to visit the Yearly Project Fair that will be held on Tuesday, June 26, 2018, starting at 13:00, in the CS Taub Lobby and 1st Floor. More details in the attached poster. You are all invtied!
[Full version]

A polyomino is an edge-connected set of cells on the square lattice. The problem of counting poyominoes dates back to the 1950s when it was studied in parallel in the fields of combinatorics and statistical physics. The study of polyominoes usually classifies polyominoes by their area, which is the number of cells the polyomino contains. Less explored is the perimeter of a polyomino, which is the set of empty cells adjacent to the polyomino. Several ...
[Full version]

In this talk I will sketch two lower bound techniques for distributed models under bandwidth restrictions. The first is via reductions from the two party communication model, which is a well known technique for achieving lower bounds in distributed computing. I will show a new gadget for this framework that allows us to prove stronger lower bounds. As an example for this gadget, I will present a new lower bound for computing the exact or ...
[Full version]

We present a new structure-preserving numerical scheme for solving the Euler–Poincaré Differential (EPDiff) equation on arbitrary triangle meshes. Unlike existing techniques, our method solves the difficult non-linear EPDiff equation by constructing energy preserving, yet fully explicit, update rules. Our approach uses standard differential operators on triangle meshes, allowing for a simple and efficient implementation.
[Full version]

Bugs, possibly leading to security flaws, are inevitable in the extremely complex modern processors. Some of such bugs may be later fixed in the field by patching processor firmware. In this presentation we will review mechanisms provided by the Intel Software Guard Extensions (SGX) architecture to recover from security vulnerabilities and to re-establish trust in the recovered platform. Bio: Ilya Alexandrovich is a Principal Engineer in the Intel Core Architecture Group. Since joining Intel eight ...
[Full version]

Collision resistant hashing is a fundamental concept that is the basis for many of the important cryptographic primitives and protocols. Collision resistant hashing is a family of compressing functions such that no efficient adversary can find {\em any} collision given a random function in the family. In this work we study a relaxation of collision resistance called \emph{distributional} collision resistance, introduced by Dubrov and Ishai (STOC '06). This relaxation of collision resistance only guarantees that ...
[Full version]

Single Image Super-Resolution (SISR) aims to recover a high-resolution image from a given low resolution version of it (the given image is assumed to be a blurred, down- sampled and noisy version of the original image). Video Super Resolution (VSR) targets series of given images, aiming to fuse them to create a higher resolution outcome. Although SISR and VSR seem to have a lot in common, as only the input domain changes between the two, ...
[Full version]

The 8th CS Research Day for graduate studies will be held on Monday, June 18, 2018 between 15:00-17:00, 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 participating researches will be on various topics: ...
[Full version]

We introduce a 3D shape generative model based on deep neural networks. A new image-like (i.e., tensor) data representation for genus-zero 3D shapes is devised. It is based on the observation that complicated shapes can be well represented by multiple parameterizations (charts), each focusing on a different part of the shape. The new tensor data representation is used as input to Generative Adversarial Networks for the task of 3D shape generation. The 3D shape tensor ...
[Full version]

The 8th annual international TCE conference on Deep Learning: Theory & Practice will take place on Thurs June 14, 2018 at the Technion Electrical Engneering Department, Meyer 280, and will focus on why is it working so well, and how can we improve it in various domains, such as vision, language, and audio. Conference Chairs: Daniel Soudry (EE Technion) and Ran El-Yaniv (CS Technion) Confirmed Speakers list includes: · Lior Wolf, Tel Aviv University, Israel ...
[Full version]

The task of rendezvous (also called gathering) calls for a meeting of two or more mobile entities, starting from different positions in some environment. Those entities are called mobile agents or robots, and the environment can be a network modelled as a graph or a terrain in the plane, possibly with obstacles. The rendezvous problem has been studied in many different scenarios. Two among many adopted assumptions particularly influence the methodology to be used to ...
[Full version]

Triangle counting is a scalable analytic that benefits from a large number of processors. Similar to many other graph analytics, the control flow for each thread of triangle counting algorithms is determined by the input graph. This data dependency makes both load-balancing and compiler optimization much more challenging. Vectorization of such codes is even more complex. The recent branch-avoiding model shows how to remove such control flow restrictions by replacing branches with an equivalent set ...
[Full version]

Over the past decade, the computer science research community has converged around a formal notion of data privacy, known as differential privacy, and has made substantial progress in establishing the theoretical foundations of this notion. In this talk, I will give a brief overview of differential privacy and the relevant mathematical toolkit, and then we will discuss the implications and frontiers of this research space. Can differential privacy ever be practical? How might it be ...
[Full version]

Dictionary Learning techniques aim to find sparse signal representations that capture prominent characteristics in the given data. For signals residing on non-Euclidean topologies, represented by weighted graphs, an additional challenge is incorporating the underlying geometric structure of the data domain into the learning process. In this talk, we introduce an approach that aims to infer and preserve the local intrinsic geometry of the data. Combining ideas from spectral graph theory, manifold learning and sparse representations, ...
[Full version]

In typical imaging systems, an image/video is first acquired, then compressed for transmission or storage, and eventually presented to human observers using different and often imperfect display devices. While the resulting quality of the perceived output image may severely be affected by the acquisition and display processes, these degradations are usually ignored in the compression stage, leading to an overall sub-optimal system performance. In this work we propose a compression methodology to optimize the system's ...
[Full version]

Covert communication considers the following problem -- if Alice wishes towhisper to Bob while ensuring that the eavesdropper Eve cannot even detect whether or not Alice is whispering, how much can she whisper. Ensuring such a stringent security requirement can be met requires new ideas from information theory, coding theory, and cryptography. In this talk I will survey the state of the existing literature (recent information-theoretic capacity-style results for a variety of settings), and then ...
[Full version]

Subdivision surfaces are a mainstream methodology in computer graphics and geometry processing to create smooth surfaces with a multiresolution hierarchy. The recent popularity of Isogeometric Analysis brought a renewed interest in such surfaces for the purpose of solving differential equations. We define subdivision methods for piecewise-constant directional fields and show how such methods can be used for robust and efficient vector field processing on subdivision surfaces.
[Full version]

Oblivious RAM (ORAM) is a cryptographic primitive that allows a client to securely execute RAM programs over data that is stored in an untrusted server. Distributed Oblivious RAM is a variant of ORAM, where the data is stored in m non-colluding servers. Extensive research over the last few decades have succeeded to reduce the bandwidth overhead of ORAM schemes, both in the single-server and the multi-server setting, from O(\sqrt{N}) to O(1). However, all known protocols ...
[Full version]

Matching pixels is used in various computer vision applications such as template matching, tracking and image editing. I will give an overview of my work in this field with an emphasize on two components. The first is how to represent the data, and the second is what similarity measure to use. I will demonstrate the results on several applications including object tracking and image editing. Bio: Shai Avidan is an Associate Professor at the School ...
[Full version]

The Technion Hiroshi Fujiwara Cyber Security Research Center is happy to invite you to the Qubit 2018 - Quantum Communication: Celebrating Bennett & Brassard's Wolf Prize for Physics conference to be held on Sunday, June 3rd, 2018 at the Taub CS Building, Technion. Chairs:Eli Biham, Technion Tal Mor, Technion Speakers will be: Keynote: Charles Bennett, IBM Research Center: "Why DIY Randomness is Better Than DI Randomness" Keynote: Gilles Brassard, Université de Montréal: "Cryptography In A ...
[Full version]

The dynamic NFV placement problem captures one of the main challenges facing the telecom industry following the emergence of the Network Function Virtualization (NFV) networking paradigm, that is, deciding the placement of network functions while taking into consideration the dynamic nature of networks and workloads. We model the problem as a generalization of the classic Uncapacitated Facility Location (UFL) problem, where we consider both multiple types of commodities and dynamic clients whose location changes over ...
[Full version]

The bulk of the talk will deal with space requirements of reliable storage algorithms in asynchronous distributed systems. A number of recent works have used codes in order to achieve a better storage cost than the well-known replication approach. However, a closer look reveals that they incur extra costs in certain scenarios. Specifically, if multiple clients access the storage concurrently, then existing asynchronous code- based algorithms may store a number of copies of the data ...
[Full version]

The availability of large idea repositories (e.g., the U.S. patent database) could significantly accelerate innovation and discovery by providing people with inspiration from solutions to analogous problems. However, finding useful analogies in these large, messy, real-world repositories remains a persistent challenge for either human or automated methods. In this work we explore the viability and value of learning simpler structural representations which specify the purpose of a product and the mechanisms by which it achieves ...
[Full version]

Image restoration algorithms are typically evaluated by some distortion measure (e.g. PSNR, SSIM, IFC, VIF) or by human opinion scores that quantify perceived perceptual quality. In this work, we prove mathematically that distortion and perceptual quality are at odds with each other. Specifically, we study the optimal probability for correctly discriminating the outputs of an image restoration algorithm from real images. We show that as the mean distortion decreases, this probability must increase (indicating worse ...
[Full version]

We are happy to invite your to the sixth of series of meetings on career and job seeking which will be held at CS, on Monay, May 28, at 17:00, in room 337, CS Taub Building. Sarah Karu (Talent management specialist) will give a talk on "How to manage negotiation on employment contract". Please pre-register. See you there!
[Full version]

We propose a new iterative algorithm for computing smooth cross fields on triangle meshes that is simple, easily parallelizable on the GPU and finds solutions with lower energy and fewer cone singularities than state-of-the-art methods. Our approach is based on a formal equivalence, which we prove, between two formulations of the optimization problem. This equivalence allows us to eliminate the real variables and design an efficient grid search algorithm for the cone singularities. We leverage ...
[Full version]

Set functions with convenient properties (such as submodularity) often arise in algorithmic game theory, and allow for improved properties of optimization algorithms and mechanisms. It is natural to ask (e.g., in the context of data driven applications) how robust such properties are, and whether small deviations from them can be tolerated. We consider two such questions in the important special case of linear set functions. One question that we address is whether any set function ...
[Full version]

Visual computing is a wide area that includes computer graphics and image processing, where the "eyeball-norm" rules. I will briefly discuss two case studies involving numerical methods and analysis applied to this area. The first case study involves motion simulation and calibration of soft objects such as plants, skin, and cloth. The governing elastodynamics PDE system, discretized in space already at the variational level using co-rotated FEM, leads to a large, expensive to assemble, dynamical ...
[Full version]

Who needs a new Instruction Set Architecture (ISA)? Architectures have reached some unspoken Truce through Markets Segment dominance (IA64 - PC and Server market, ARM - Mobile market). IA and ARM have ruled the PC/Server and Mobile markets, respectively, for years and have prevailed aggressive competitive assaults Some Markets are still pursuing a standard (ARM and Various DSPs - IoT market, GPGPU (Nvidia) and TPU (Google) - IA/ML applications) IoT and IA/ML segments have not ...
[Full version]

We present a novel approach to template matching that is efficient, can handle partial occlusions, and comes with provable performance guarantees. A key component of the method is a reduction that transforms the problem of searching a nearest neighbor among N high-dimensional vectors, to searching neighbors among two sets of order sqrt{N} vectors, which can be found efficiently using range search techniques. This allows for a quadratic improvement in search complexity, and makes the method ...
[Full version]

Modern software systems are expected to be secure and contain all the latest features, even when new versions of software are released multiple times an hour. Each system may include many interacting packages. The problem of installing multiple dependent packages has been extensively studied in the past, yielding some promising solutions that work well in practice. However, these assume that the developers declare all the dependencies and conflicts between the packages. Often, the entire repository ...
[Full version]

Computer systems have been developed tremendously over the past few years and as a result new challenges arise in the fields of security and reliability. In this talk I will present two approaches that utilize the operating system for providing solutions to urgent challenges in these fields. First, I will speak about the Core Surprise Removal (CSR) algorithm which explores the uncharted field of providing reliability in face of hardware faults and makes a unique ...
[Full version]

Recent progress in imaging technologies leads to a continuous growth in biomedical data, which can provide better insight into important clinical and biological questions. Advanced machine learning techniques, such as artificial neural networks are brought to bear on addressing fundamental medical image computing challenges such as segmentation, classification and reconstruction, required for meaningful analysis of the data. Nevertheless, the main bottleneck, which is the lack of annotated examples or ‘ground truth’ to be used for ...
[Full version]

Oblivious RAM (ORAM) is a cryptographic primitive that allows a client to securely execute RAM programs over data that is stored in an untrusted server. Distributed Oblivious RAM is a variant of ORAM, where the data is stored in m non-colluding servers. Extensive research over the last few decades have succeeded to reduce the bandwidth overhead of ORAM schemes, both in the single-server and the multi-server setting, from O(\sqrt{N}) to O(1). However, all known protocols ...
[Full version]

In digital geometry, Euclidean objects are represented by their discrete approximations e.g., subsets of the lattice of integers. Rigid motions of such sets have to be defined as maps from and onto a given discrete space. One way to design such motions is to combine continuous rigid motions defined on Euclidean space with a digitization operator. However, digitized rigid motions often no longer satisfy properties of their continuous siblings. Indeed, due to digitization, such transformations ...
[Full version]

The review will include the main modern cyber security challenges that exists in vehicles today and in the coming future. There will specific focus on security aspects in vehicle-to-vehicle communication (v2v or C2C) , includes privacy , authenticity and integrity of the communication in the relevant scenarios Bio: Yoni is managing the product Cyber Security Group in General Motors Israel that in-charge to secure crucial elements in the vehicle. Yoni was the chairman of the ...
[Full version]

In extreme classification problems, machine learning algorithms are required to map instances to labels from an extremely large label set. We build on a recent extreme classification framework with logarithmic time and space, and on a general approach for error correcting output coding (ECOC), and introduce a flexible and efficient approach accompanied by bounds. Our framework employs output codes induced by graphs, and offers a tradeoff between accuracy and model size. We show how to ...
[Full version]

A recent principled approach to information extraction from text views the query as an ordinary relational query, but not on ordinary relations; instead, the relations are tuples of intervals, or "spans," extracted from text documents. A prominent example is Fagin et al.'s "regular spanners," which are the closure of regular expressions with capture variables (used for tokenization, n-grams, segmentation, etc.) under the relational algebra. In this talk I will present several research directions motivated by ...
[Full version]

In this talk we will present improvements to the accuracy, memory and speed of counting algorithms. Counters are one of the most basic building blocks in networking, big data analytics and other fields. To cope with increasing line rates, counters must be accessed quickly and be kept on small fast memory. Further, to reach concise and exhaustive conclusions about data, accurate estimations must be kept for large numbers of counters. We first present a closed ...
[Full version]

As the first and most popular decentralized cryptocurrency to date, Bitcoin has ignited much excitement, not only for its novel realization of a central bank-free financial instrument, but also as an alternative approach to classical problems in distributed computing and cryptographic protocols, such as reaching consensus in the presence of misbehaving parties. In this talk, after a brief introduction to the innovative and distributedly-maintained data structure known as the "blockchain," we first present the first ...
[Full version]

CS is happy to invite you to the third of series of workshops that will hand you applicable tools for working, developing and managing code projects, and this time on: "Workin with Distance Servers: How to Submit Homework without Sleeping in the Computer Farm", by Yonit Gruber-Hazani, DEVOPSE engineer in WAZE. In the program: - Tips for working from the command line - How to connect to other computers, and why it is recommended not ...
[Full version]

In recent years, improvements in various scientific image acquisition techniques gave rise to the need for adaptive processing methods, particularly aimed for large data-sets corrupted by noise and deformations. Motivated by challenges in cryo-electron microscopy (cryo-EM), we consider the problem of reducing noise in a dataset of images admitting a certain unified structure. In particular, we consider datasets of images sampled from an underlying low-dimensional manifold (i.e. an image-valued manifold), where the images are obtained ...
[Full version]

CS invites you to assemble a winning team and start working on a brilliant idea for the Spring Hackaton that will take place on Thursday-Friday, May 3-4, 2018 in CS Taub Building. The program will include 24 hour team work, supervised by leading tutors, refreshments and pecuniary rewards (first place will win 5,000 NIS). Please pre-register by April 15th, seats are limited. See you there!
[Full version]

We present a new model of probabilistically checkable proof (PCP), which we call "Distributed PCP": A satisfying assignment (x in {0,1}^n) to a SAT instance is shared between two parties (Alice knows x_1, ... x_{n/2}, and Bob knows x_{n/2+1},...,x_n). Their goal is to jointly write a PCP for x, while exchanging little or no information. Using our new framework, we obtain, for the first time, PCP-like hardness of approximation results for problems in P. Based ...
[Full version]

In the last decade, the human population has produced zettabytes (10^21) of digital data. This creates immense opportunities and challenges for biology research. In this talk, I will present two research directions on the intersection between genetics and data, which we dub "genetic media". First, I will speak about crowd sourcing massive genetic data using social media. We collected over 80 million profiles from the largest social-media website driven by genealogy and constructed a single ...
[Full version]

We propose a novel measure for template matching named Deformable Diversity Similarity -- based on the diversity of feature matches between a target image window and the template. We rely on both local appearance and geometric information that jointly lead to a powerful approach for matching. Our key contribution is a similarity measure, that is robust to complex deformations, significant background clutter, and occlusions. Empirical evaluation on the most up-to-date benchmark shows that our method ...
[Full version]

We are happy to invite your to the fifth of series of meetings on career and job seeking which will be held at CS, on Monay, April 30, at 17:00, in room 337, CS Taub Building. Mr. David Gross, Program Engineer in Facebook Israel will give a talk on how to prepare and succeed in a technical interview: "Crush Your Coding Interview". Please pre-register. See you there!
[Full version]

EUROCRYPT 2018 is the 37th Annual International Conference on the Theory and Applications of Cryptographic Techniques. Covering all aspects of cryptology, including theoretical foundations, deployment of cryptographic schemes, cryptanalysis of widely used standards, cryptographic protocols, quantum cryptography, and cryptographic currencies. This year's edition will take place in Tel Aviv, Israel on April 29-May 3 2018. EUROCRYPT 2018 is one of the flagship conferences of the the International Association for Cryptologic Research (IACR), and is organized by the Technion Hiroshi Fujiwara Cyber Security Research ...
[Full version]

We are happy to invite you to a special meeting fot graduate students that will deal this time with research career: is there research in Israeli industry and what are the current available jobs and career tracks. The event will be held at CS, on Sunday, April 29rd, at 15:30, in the Grads Club, floor 2, CS Taub Building. In the program: Prof. Danny Raz Technion CS   Michal Aharon Sr. director research, Oath (Yahoo) ...
[Full version]

A graph learning problem is a problem of finding a hidden graph $G=(V,E)$ using edge-detecting queries, where an edge-detecting query $Q_G(S)$, for $S \subseteq V$ is: does S contain at least one edge of G? The main question is how many queries do we need to find all the edges. Graph learning is a well-studied problem. It has been studied for general graphs, and also for specific graph families (i.e. matching, stars, cliques and others). ...
[Full version]

A graph learning problem is a problem of finding a hidden graph $G=(V,E)$ using edge-detecting queries, where an edge-detecting query $Q_G(S)$, for $S \subseteq V$ is: does S contain at least one edge of G? The main question is how many queries do we need to find all the edges. Graph learning is a well-studied problem. It has been studied for general graphs, and also for specific graph families (i.e. matching, stars, cliques and others). ...
[Full version]

A k-spanner is a sparse subgraph that preserves distances up to a multiplicative factor of k. First introduced in the late 80's, spanners have been central for numerous applications, such as synchronization, compact routing tables, approximate shortest paths, and more. This talk focuses on distributed construction of approximate minimum spanners, presenting both algorithmic and hardness results, in the classical LOCAL and CONGEST models of distributed computing. In both models vertices exchange messages in synchronous rounds, ...
[Full version]

The 2018 open day invites outstanding undergraduates from all universities to come to the Technion and learn about the Computer Science and Department, to meet faculty and graduate students and to hear a fascinating talk by Moshe Levinger, Senior Manager in IBM Haifa: "Do Advanced Degrees Indeed Advance?" The event will be held on Wednesday, April 25, 2018, between 12:00, in CS Taub Building, room 337 (3rd floor). The program will include review on curriculum and ...
[Full version]

AI, Cognitive, Analytics, IoT all run on data. This data is needed by data scientists, line of business, developers, CDOs and others. Without data, in fact, one can do very little that is interesting or of value. Object storage, the default storage for the PBs and EBs of unstructured data in clouds, has brought huge efficiencies to storing and managing data. In this presentation, we will provide an overview of object storage, using IBM's Cloud ...
[Full version]

GANs have recently gained attention due to their success in generating realistic new samples of natural images, yet the extent to which such models capture the statistics of full images is poorly understood. In this work we present a simple method to evaluate generative models based on relative proportions of samples that fall into predetermined bins. Applying our method to GANs shows that they typically fail to capture very basic properties of the distribution. As ...
[Full version]

CS is happy to invite you to The Marconi Society conference: "The Information Era: Past, Present and Future", on Tuesday, April 24th, 2018 in Taub Auditorium 2. Guest of honor will be Prof. Andrew Viterbi, one of Qualcomm founders, who will take part in a historic panel with Prof. Jacob Ziv and Prof. Abraham Lempel, in celebration of 70 years of information and communication. The conference is in cooperation with Marconi Society, Haifa Municipality and ...
[Full version]

Lattice animals are connected subgraphs of a lattice. In this talk, we consider two types of lattices, the $d$-dimensional hypercubic lattice and the triangular lattice, where the animals are frequently referred to as $d$-dimensional polycubes and polyiamonds, respectively. Denote the number of $d$-dimensional polycubes of volume $n$ and perimeter defect $k$ as $B(n,k,d)$, where $k$ is the deviation from the maximum perimeter. To-date, no formulae of $B(n,k,d)$ are known. In the first part of the ...
[Full version]

Labeling textual data by humans can become very time-consuming and expensive, yet critical for the success of an automatic text classification system. In order to minimize the human-labeling efforts, we propose a novel active learning (AL) solution, which does not rely on existing sources of unlabeled data. It uses a small amount of of labeled data as the core set for the synthesis of useful membership queries (MQs) - unlabeled instances synthesized by a computer, ...
[Full version]

Future servers are expected to have huge amount of volatile and nonvolatile main memory that can be directly accessed by the processors. Managing such a memory hierarchy is a major concern of computer and of hardware security architects. Thus, while computer architects are aiming at increasing the page and region size, which is being controlled by a single entry at of the TLB (e.g., 1G page size), hardware security architects aims at protecting variable size ...
[Full version]

Given a set of points (sites) in the plane, a Voronoi diagram is a partitioning of the plane into regions such that each region contains one site and all points closer to this site than to any other site. Voronoi diagrams have practical and theoretical applications in a large number of fields. In this talk, I will overview the exciting new uses of Voronoi diagrams for planar graph algorithms. In particular, I will describe an ...
[Full version]

This thesis addresses the problem of viewpoint estimation of an object in a given image, where the objects belong to several known categories. Convolutional Neural Networks were recently applied to this problem, leading to large improvements of state-of-the-art results. Two major approaches have been pursued: a regression approach, which handles the continuous values of view points naturally, and a classification approach, which discretized the space of viewpoints. We follow the second approach and present five ...
[Full version]

A MAGIC SQUARE is an arrangement of nxn disjoint integers on an nxn square in a way such that the n rows, the n columns and the two diagonals of the square sum to the same constant which is called THE MAGIC CONSTANT. We will consider in this lecture magic squares with n=3 only. We introduce a new generalization of magic squares where 6 non equivalent positive magic squares having the same magic constant are ...
[Full version]

The memory interface is a striking example of a hardware/software interface that is ill-suited for parallelism, and specifically for parallel access to the program's state. Even on modern computers, that interface follows the design described in the von Neumann architecture more than 70 years ago. This interface design is fundamentally sequential and suited for serial instruction execution. Indeed, modern memory systems allow multiple cores to read and write state concurrently, and the underlying implementation can ...
[Full version]

Video scene detection is the task of temporally dividing a video into its semantic sections called "scenes" - a series of video shots depicting a high-level concept or story (action/drama scene, news segment etc.). This is an important preliminary step for effective analysis of heterogeneous video content, and can help with building a table-of-contents, enabling fast browsing and skipping between scenes, and identifying the contextual boundaries of the content in the video. In this talk ...
[Full version]

CS is happy to invite you to the second of series of workshops that will hand you applicable tools for working, developing and managing code projects, and this time on: "How to work together on a code project (and to stay friends)": - How to divide the work when there is no time to meet? - What to do when one partner is weaker/stronger than the others? - How to work with four-hand and one ...
[Full version]

A recent result shows that the Java type system and its type-checker are mighty enough to emulate deterministic pushdown automata (DPDAs), and hence, capable (disregarding cost) to recognize deterministic context free languages (DCFGs). The problem is of concrete practical value, since, as it turns out, this recognition is essential for automatic generation of fluent APIs from their specification. This work advances the state of knowledge, in presenting, for the first, an efficient (specif- ically linear ...
[Full version]

The quality of unit tests gains substantial importance in modern software systems. We explore the way in which Junit tests are written in real world Java systems. We analyse 112 Java repositories and measure the quality of unit tests by finding patterns which indicate good practices of coding. Our results show that the quality of real world unit tests is low, and that in many cases, unit tests don’t follow the well-known recommendations for writing ...
[Full version]

High-availability requirements in modern software triggered the widespread adoption of eventually consistent data stores. Unfortunately, reasoning about the correctness of programs running under eventual consistency is a challenging task, largely due to the possibility of very weak system behaviors. In this talk, I will present new automated techniques that help developers precisely identify problematic weak behaviors. Our starting point will be to adopt conflict serializability as a correctness criterion. We then give a new generalized ...
[Full version]

Deep Learning has led to a dramatic leap in Super-Resolution (SR) performance in the past few years. However, being supervised, these SR methods are restricted to specific training data, where the acquisition of the low-resolution (LR) images from their high-resolution (HR) counterparts is predetermined (e.g., bicubic downscaling), without any distracting artifacts (e.g., sensor noise, image compression, non-ideal PSF, etc). Real LR images, however, rarely obey these restrictions, resulting in poor SR results by SotA (State ...
[Full version]

Large-scale storage systems lie at the heart of the big data revolution. As these systems grow in scale and capacity, their complexity grows accordingly, building on new storage media, hybrid memory hierarchies, and distributed architectures. Numerous layers of abstraction hide this complexity from the applications, but also hide valuable information that could improve the system's performance considerably. I will demonstrate how to bridge this semantic gap in the context of erasure codes, which are used ...
[Full version]

eBay will hold a special event designated for graduate students on Thursday, March 22, 2018, between 10:30-12:00, in room 337, CS Taub Building. The program includes opening by Ido Guy, eBay Israel Head of Research, a lecture by Dr. Kira Radinsky, eBay Israel Chief Scientist, and Q&A sesssion. Full program and more details in the attached poster. Attendance requires pre-registration. You are all invited!
[Full version]

You are invited to the final stage of the 2018 Best Project Contest. The finalist teams will present and talk about their projects. The event will take place on Wednesday, March 21, 2018, between 12:30-15:00, in Auditorium 2, CS Taub Building. You are all invited to cheer, support and to watch the most outstanding CS projects.
[Full version]

In the classical firing squad problem, an unknown number of nodes represented by identical finite states machines is arranged on a line and in each time unit each node may change its state according to its neighbors' states. Initially all nodes are passive, except one specific node located at an end of the line, which issues a fire command. This command needs to be propagated to all other nodes, so that eventually all nodes simultaneously ...
[Full version]

We introduce a new approach for identifying individuals based on their motion patterns in interactive scenarios. We formalize the identification process in the context of a sequential message exchange session between the subject and the system. The subject is modeled with a probabilistic generative model inspired by the Human Information Processing (HIP) paradigm. At each stage, the system presents a visual stimulus (a cue) to the subject and records their motion response. The cue is ...
[Full version]

We introduce a new approach for identifying individuals based on their motion patterns in interactive scenarios. We formalize the identification process in the context of a sequential message exchange session between the subject and the system. The subject is modeled with a probabilistic generative model inspired by the Human Information Processing (HIP) paradigm. At each stage, the system presents a visual stimulus (a cue) to the subject and records their motion response. The cue is ...
[Full version]

Research in medical image analysis, especially after the deep-learning revolution, relies on access to medical image databases. Specific categories of medical images have been made available as “challenges” to facilitate scientific benchmarking. Nevertheless, R&D of real-world clinical technologies usually requires data acquired while diagnosing and treating particular classes of patients using specific protocols and equipment. Such data cannot be normally found in common challenges, and can only be obtained from hospitals and health care providers. ...
[Full version]

We consider a multi-level aggregation problem in a weighted rooted tree, studied recently by Bienkowski et al. In this problem requests arrive over time at the nodes of the tree, and each request specifies a non-decreasing waiting costs. A request is served by sending it to the root before its deadline at a cost equal to the weight of the path from the node in which it resides to the root. However, requests from different ...
[Full version]

We unveil the existence of a vulnerability in Wi-Fi (802.11) networks, which allows an adversary to remotely launch a Denial-of- Service (DoS) attack that propagates both in time and space. This vulnerability stems from a coupling effect induced by hidden nodes. Cascading DoS attacks can congest an entire network and do not require the adversary to violate any protocol. We demonstrate the feasibility of such attacks through experiments with real Wi-Fi cards and extensive ns-3 ...
[Full version]

Quad-Copters are versatile unmanned aircraft that are used for a myriad of tasks from manual aerial photography to autonomous surveillance. After their initial success in outdoors scenarios, indoor autonomous activity has recently been getting attention in research. Unlike the outdoor environment, which offers cheap and easy localization in the form of GPS, and usually lacks meaningful obstacles, indoor autonomous flight lacks a common means of localization and the tight spaces are fraught with every day ...
[Full version]

Quad-Copters are versatile unmanned aircraft that are used for a myriad of tasks from manual aerial photography to autonomous surveillance. After their initial success in outdoors scenarios, indoor autonomous activity has recently been getting attention in research. Unlike the outdoor environment, which offers cheap and easy localization in the form of GPS, and usually lacks meaningful obstacles, indoor autonomous flight lacks a common means of localization and the tight spaces are fraught with every day ...
[Full version]

Quad-Copters are versatile unmanned aircraft that are used for a myriad of tasks from manual aerial photography to autonomous surveillance. After their initial success in outdoors scenarios, indoor autonomous activity has recently been getting attention in research. Unlike the outdoor environment, which offers cheap and easy localization in the form of GPS, and usually lacks meaningful obstacles, indoor autonomous flight lacks a common means of localization and the tight spaces are fraught with every day ...
[Full version]

Quad-Copters are versatile unmanned aircraft that are used for a myriad of tasks from manual aerial photography to autonomous surveillance. After their initial success in outdoors scenarios, indoor autonomous activity has recently been getting attention in research. Unlike the outdoor environment, which offers cheap and easy localization in the form of GPS, and usually lacks meaningful obstacles, indoor autonomous flight lacks a common means of localization and the tight spaces are fraught with every day ...
[Full version]

Quad-Copters are versatile unmanned aircraft that are used for a myriad of tasks from manual aerial photography to autonomous surveillance. After their initial success in outdoors scenarios, indoor autonomous activity has recently been getting attention in research. Unlike the outdoor environment, which offers cheap and easy localization in the form of GPS, and usually lacks meaningful obstacles, indoor autonomous flight lacks a common means of localization and the tight spaces are fraught with every day ...
[Full version]

In 2015, Intel launched its 6th generation Core, codenamed Skylake, that implements a new ISA for security – Intel® SGX. The talk will provide a short intro to SGX for the unfortunates who have not been exposed to Intel® SGX yet, but will focus mainly around the architectural enhancements since its first introduction and some of the design choices that were made. Short Bio: Ittai Anati is a Senior Principal Engineer at Intel Corporation’s CPU ...
[Full version]

Google will host a hub at CS for the 2018 Online Qualification Round of Hash Code on Thursday, March 1st, at 18:30 CET, in CS Taub 337. Participants are requested to pre-.
[Full version]

Using selective regression, it is possible to increase accuracy of predictions by abstaining from answering when there is insufficient knowledge. This work is about increasing the accuracy of selective regression even further and using simple selective models to create a more complex one, by using an ensemble of selective regressors. We demonstrate how to achieve improved accuracy by using two methods to build our ensemble. In the first approach, we first split the samples in ...
[Full version]

Frequency modulation (FM) is a form of radio broadcasting which is widely used nowadays and has been for almost a century. The widest use of FM is for radio broadcasting, which is commonly used for transmitting audio signal representing voice. Due to the effect of various distortions, noise conditions and other impairments imposed on the transmitted signal, the detection reliability severely deteriorates. As a result thereof, the intelligibility and quality of the detected speech decreases ...
[Full version]

Trying to predict which items are likely to be accessed in the near future is the basis for most cache management policies. Storage workloads are often characterized by their level of frequency bias and their level of recency bias. The former captures how well the access frequency serves as a predictor for future accesses while the latter indicates how well does access recency predicts the future. Existing cache management policies have been grappling with the ...
[Full version]

The direct memory access (DMA) mechanism allows I/O devices to independently access the memory without CPU involvement, improving performance but exposing systems to malicious DMA attacks. Hardware vendors therefore introduced IOMMUs (I/O memory management units), allowing operating systems to defend themselves by restricting DMAs to specific memory locations. When configured correctly, the latest generation of IOMMUs is thus considered an appropriate solution to the problem. We challenge this perception and uncover a new type of ...
[Full version]

In modern cloud infrastructures, each physical server often runs multiple virtual machines and employs a software Virtual Switch (VS) to handle their traffic. In addition to switching, the VS performs network measurements, such as identifying the most frequent flows, which are essential for networking applications such as load balancing and intrusion detection. Unlike traditional streaming algorithms, which minimize the space requirements, the bottleneck in virtual switching measurement is the CPU utilization. In this talk, I ...
[Full version]

We present novel probabilistic gathering algorithms for agents that can only detect the presence of other agents in front or behind them. The agents act in the plane and are identical and indistinguishable, oblivious and lack any means of direct communication. They do not have a common frame of reference in the plane and choose their orientation (direction of possible motion) at random. The analysis of the gathering process assumes that the agents act synchronously ...
[Full version]

The surface separating the interior of a given object or body from its exterior defines a shape. A key challenge in computer vision is to recover the shape from missing, partial, sparse or noisy observations such as images, depth maps or position sensors. Although in theory, surfaces can be arbitrarily embedded in space, the underlying structure of deformable shapes, such as the human face, is often driven by just a few degrees of freedom. For ...
[Full version]

Textual data written in some natural language carries concealed and valuable information within. Information Extraction (IE) is the task of automatically extracting this information in a structured representation. Standard relational database systems, who are highly suitable for representing structured information, are in fact incapable of performing deep text analysis, and therefore out-of-database solutions are often applied. However, this approach is prone to laborious development processes, complex and tangled programs, and inefficient control flows. These deficiencies ...
[Full version]

The rapid digitization of genealogical and medical records enables the assembly of extremely large pedigree records spanning millions of individuals. Such pedigrees provide the opportunity to answer genetic and epidemiological questions in scales much larger than previously possible. Linear mixed models (LMMs) are often used for analysis of pedigree data. However, LMMs cannot naturally scale to large pedigrees spanning millions of individuals, owing to their steep computational and storage requirements. Here we propose a novel ...
[Full version]

When sensitive data is stored in the cloud, the only way to ensure its secrecy is by encrypting it before it is uploaded. Not only is encryption itself computationally expensive but the encryption keys must also be safely stored and the data decrypted whenever it is downloaded. The emerging multi-cloud model, in which data is stored redundantly in two or more independent clouds, provides an opportunity to protect sensitive data with secret-sharing schemes. This approach ...
[Full version]

You are invited to a special lecture and ceremony celebrating 40th Anniversary of the Lempel-Ziv Algorithm, by Prof. Meir Feder from the School of Electrical Engineering in Tel-Aviv University on: "Beyond Compression: Lempel-Ziv in Learning and Prediction". The event will take place on Monday, February 5th, 2018, at 14:00, in the Auditorium 1003, EE Meyer Building, Technion. Program: 14:00 - Refreshments 14:30 - Celebration 15:00 - Lecture More details. You are all invited.
[Full version]

How does computational learning change when one cannot store all the examples one sees in memory? This question has seen a burst of interest in the past couple of years, leading to the surprising theorem that there exist simple concepts (parities) that require an extraordinary amount of time to learn unless one has quite a lot of memory. In this work we show that in fact most concepts cannot be learned without sufficient memory. This ...
[Full version]

A tree decomposition of a graph facilitates computations by grouping vertices into bags that are interconnected in an acyclic structure; hence their importance in a plethora of problems such as query evaluation over databases and inference over probabilistic graphical models. The relative benefit from different tree decompositions is measured by diverse (sometime complex) cost functions that vary from one application to another. For generic cost functions like width and fill-in, an optimal tree decomposition can ...
[Full version]

Secure system design should be guided by two principles: (1) system security should not impede third-party developers, who are often the main source of innovation, and (2) systems that secure third-party extensions also improve security by reducing the amount of specially-privileged first-party code. Unfortunately, very few systems today adhere to these principles. This is not merely a result of poor system building. It is hard to design highly extensible systems that are both secure and ...
[Full version]

We consider the image classification problem using deep models. Most of the works in the recent years consider only the flat precision (FP) measure as a benchmark. We follow a work done in recent years (called DeVISE), where a new measure named Hierarchical Precision (HP) is defined and used to measure the semantic accuracy of a classification model given an underlying hierarchy. While in DeVISE and other following works an extra side-information (e.g. textual corpus) ...
[Full version]

A set of vertices S resolves a graph if every vertex is uniquely determined by its vector of distances to the vertices in S. The metric dimension of a graph is the minimum cardinality of a resolving set of the graph. Fix a connected graph G on q ≥ 2 vertices, and let M be the distance matrix of G. We prove that if there exists w ∈ Z q such that \sum_i w_i =0 ...
[Full version]

Many problems in geometry processing, graph theory, and machine learning involve optimizations whose variables are defined over a geometric domain. The geometry of the domain gives rise to geometric structure in the optimization problem itself. In this talk, I will show how leveraging geometric structure in the optimization problem gives rise to efficient and stable algorithms applicable to a variety of application domains. In particular, I will describe new methods for problems arising in shape ...
[Full version]

Alterations in metabolic activity in tumors provide novel means to selectively target cancer cells. A powerful tool for identifying genes essential for cancer cell proliferation and survival is genome-scale RNAi and CRISPR-based genetic silencing screens. Integration of the measured gene essentiality datasets with genomic characterization of genes was shown to provide mechanistic understanding of tumor-specific gene essentiality. Here, we analyze the essentiality of metabolic enzyme-coding genes in cancer by utilizing measurements from recent large-scale genetic ...
[Full version]

While deep learning methods in Natural Language Processing are arguably overhyped, recurrent neural networks (RNNs), and in particular gated recurrent networks like the LSTM, 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 and giving a birds-eye overview of our work, I will describe a line of work in which we use LSTM encoders in a multi-task learning scenario. ...
[Full version]

The CS Computer Communication Lab (LCCN) invites you to an exposure evening to communication networks - to get to know the laboratory activity, the research and the projects it offers and the team that leads them: - Prof. Danny Raz - Laboratory Head: The Net - Not What You Thought - Itzik Ashkenazi - Lab Engineer: Network Components Programmed in P4 and special guest: - Daniel Bar-Lev - Director, MEF CTO: The Third Network - ...
[Full version]

A q-Steiner System S_q(t,k,n) is set S of k-subspaces in a space of dimension n over a finite field, such that each t-subspace of the same space is contained in exactly one subspace from S. They have applications in random network Coding, but except for one set of parameters for which t>1, their existence is unknown. Specifically, the q-Fano plane S_q(2,3,7) is the smallest structure for which its existence is unknown. In the seminar I ...
[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 to their joint course with Intel and Rafael: Virtual and Augmented Reality. The event will be held on Wednesday, January 28, 2017, between 12:30-14:30 at the CS Taub Lobby. You are all invited!
[Full version]

We study discrete-log algorithms that use preprocessing. In our model, an adversary may use a very large amount of precomputation to produce an "advice" string about a specific group (e.g., NIST P-256). In a subsequent online phase, the adversary's task is to use the preprocessed advice to quickly compute discrete logarithms in the group. Motivated by surprising recent preprocessing attacks on the discrete-log problem, we study the power and limits of such algorithms. In particular, ...
[Full version]

Cyber attacks on critical cyber systems are not decreasing in frequency or complexity. Aggressors choose the time and place of these engagements; protectors must identify, research and develop defensive techniques that provide an asymmetric advantage. A static, data-driven, preventative, automated defense is a losing strategy; an effective defense must be dynamic, behavioral, responsive and capitalize on a human in the loop. We propose human and machine performed linkography to detect, correlate, attribute and predict attacker ...
[Full version]

On Tuesday, January 23, 2018, between 12:30-14:30, at the CS Taub Loby, 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! Details on the presenting projects in the Hebrew page
[Full version]

Expressive efficiency refers to the relation between two architectures A and B, whereby any function realized by B could be replicated by A, but there exists functions realized by A, which cannot be replicated by B unless its size grows significantly larger. For example, it is known that deep networks are exponentially efficient with respect to shallow networks, in the sense that a shallow network must grow exponentially large in order to approximate the functions ...
[Full version]

Non-volatile memory is expected to coexist with (or even displace) volatile DRAM for main memory in upcoming architectures. This has led to increasing interest in the problem of designing and specifying durable data structures that can recover from system crashes. Data structures may be designed to satisfy stricter or weaker durability guarantees to provide a balance between the strength of the provided guarantees and performance overhead. This seminar proposes three novel implementations of a concurrent ...
[Full version]

A private information retrieval (PIR) protocol allows a user to retrieve a data item from a database without revealing any information about the identity of the data item to a certain coalition of servers. In this talk, we will go over the recent results on PIR, especially on PIR in MDS-coded databases with colluding servers. We will also mention various other PIR models including PIR with robust/Byzantine servers or PIR with arbitrary collusion patterns.
[Full version]

We address the problem of constructing a minimum height decision tree of a class C in polynomial time. This problem has many interesting applications that include, to name a few, computer vision, group testing, exact learning from membership queries, and game theory. We further study the combinatorial measure, the extended teaching dimension, ETD(C) of a class C. We show an algorithm that achieves a ETD(C)-approximation of the optimal height. When the extended dimension is small, ...
[Full version]

The triangle removal lemma, proved by Ruzsa and Szemerédi in 1976, states that if a graph contains a small number of triangles then it can be made triangle-free by a small number of edge deletions. The triangle removal lemma found applications in several areas of computer science and mathematics. In a series of works, culminating in a result of Alon and Shapira from 2005, it was shown that in fact, any hereditary graph property P ...
[Full version]

The security of the Bitcoin system is based on having a large amount of computational power in the hands of honest miners. Such miners are incentivized to join the system and validate transactions by the payments issued by the protocol to anyone who creates blocks. As new bitcoins creation rate decreases (halving every 4 years), the revenue derived from transaction fees start to have an increasingly important role. We argue that Bitcoin's current fee market ...
[Full version]

The Technion's Entrepreneurship Center, and the Electrical Engineering and Computer Science Departments invite you to a guest lecture by Marius Nacht, one of the founders of Check Point and chairman of the company, who has been actively investing in biomed projects, including pharmaceuticals, medical devices and digital health services. Nacht will share his rich experience and talk about why entrepreneurship and the first steps in this world. Students from all faculties for all degrees are ...
[Full version]

Recommender systems are first and foremost about matching users with items the systems believe will delight them. The "main street" of personalization is thus about modeling users and items, and matching per user the items predicted to best satisfy the user. This holds for both collaborative filtering and content-based methods. In content discovery engines, difficulties arise from the fact that the content users natively consume on publisher sites does not necessarily match the sponsored content ...
[Full version]

The composition of the earth's atmosphere has a vital effect on life on earth. Accordingly, the task of monitoring its contents, i.e. aerosol and cloud concentrations, draws much interest in science and engineering. Currently, the methods employed in monitoring the atmosphere require expensive equipment, retrieve only partial data or use simplified models (e.g. horizontally uniform atmosphere). To address these shortcomings, in this thesis, we propose both a novel system and new algorithms to sense and ...
[Full version]

In unsupervised domain mapping, the learner is given two unmatched datasets A and B. The goal is to learn a mapping G_AB that translates a sample in A to the analog sample in B. Recent approaches have shown that when learning simultaneously both G_AB and the inverse mapping G_BA, convincing mappings are obtained. In this work, we present a method of learning G_AB without learning G_BA. This is done by learning a mapping that maintains ...
[Full version]

Counting network flows’ statistics is at the heartof network monitoring, network security, and similar networkfunctionalities.Virtualization in datacenters as well as networkfunction virtualization (NFV) trends push traditional hardwareimplementations intovirtual switchesrunning as software onthe host’s CPU.The problem is that Bloom filter based datastructures for maintaining flows’ statistics such as count-minsketch and spectral Bloom filter are not optimal for softwareimplementations. TinyTable is a recent alternative which wasshown to be better adapted for being run in software, as itonly ...
[Full version]

You are invited to scholarship award ceremony for CS and EE excellent female students by Intel and DELL-EMC. The event will be held on Thursday, January 11, 2018, in room 337 (3rd floor) of CS Taub Building at the Technion. More details and program in the attached ad. You are all invited!
[Full version]

Web extraction is an important research topic that has been studied extensively, receiving a lot of attention and focus. Large amounts of data are produced and consumed online in a continuous and growing rate. The ability to collect and analyze these data has become essential for enabling a wide range of applications and improving the effectiveness of modern businesses. Web extraction methods facilitate the collection and analysis of these data by transforming the human friendly ...
[Full version]

In this work we consider the problem of testing whether a graph has bounded arboricity. The family of graphs with bounded arboricity includes, among others, bounded-degree graphs, all minor-closed graph classes (e.g. planar graphs, graphs with bounded treewidth) and randomly generated preferential attachment graphs. Graphs with bounded arboricity have been studied extensively in the past, in particular since for many problems they allow for much more efficient algorithms and/or better approximation ratios. We present a ...
[Full version]

Caching is the most effective practice for reducing delays and congestion in networks aimed at content distribution. In this work we address the setup of multiple Internet service providers (ISPs) sharing a federated cache. When multiple ISPs with different demands share a cache, storage cost can be reduced when the cache is coded, that is, storing objects that are XORs of multiple content objects. The talk will present our results on low-complexity methods for coded ...
[Full version]

In recent years, robots have played an active role in everyday life: medical robots assist in complex surgeries, low-cost commercial robots clean houses and fleets of robots are used to efficiently manage warehouses. A key challenge in these systems is motion planning, where we are interested in planning a collision-free path for a robot in an environment cluttered with obstacles. While the general problem has been studied for several decades now, these new applications introduce ...
[Full version]

Increasingly adopted Web-based knowledge sources, such as knowledge graphs and crowdsourcing platforms, pose a great challenge in making them accessible to non-expert end-users. In particular, being unaware of the properties, structure and contents of available knowledge sources, end-users may not know which sources to use and in what manner. In this talk I will present several directions towards bridging this gap. First, we assist users in identifying adequate knowledge sources, via a declarative framework that ...
[Full version]

In several clinical routines such as mammography, the early detection of breast cancer has an enormous impact in the patient survival. Radiologists nowadays are overwhelmed by the imaging data coming from numerous screening and diagnostic tools. This puts pressure on the early detection of breast cancer has an enormous impact in the patient survival. Radiologists nowadays are overwhelmed by the imaging data coming from numerous screening and diagnostic tools. This puts pressure on the radiologists ...
[Full version]

Computational problems whose input is a program are central in Cryptography, as well as Complexity, Learning, and Optimization. The nature of such problems crucially depends on the way the program is accessed -- as a black box or explicitly by its implementation. In which settings can we exploit code to gain an advantage over black-box access? In Cryptography, we explore this question from two opposing perspectives: - Protecting Code: Can we obfuscate a program's code ...
[Full version]

In the previous research, people tend to optimize the sensing matrix to improve the signal reconstruction accuracy for compressive sensing (CS) system. However, they assume the signal is exactly sparse which is not true in practice. So we try to find a way to design a sensing matrix which is robust to the case when the signal is not exactly sparse. Moreover, we also take the complexity of sensing a signal into account in designing ...
[Full version]

Yahoo will hold the annual event at Technion CS on Monday, January 8, 2018, 12:00-14:00, in room 337 (3rd floor) of the CS Taub Building The program includes Research Talk, Round Tables, Delightful Snacks and Surprises. Full program and more details in the attached ad. Your are all invited!
[Full version]

Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. We propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). ...
[Full version]

We address the problem of extracting an automaton from a trained recurrent neural network (RNN). We present a novel algorithm that uses exact learning and abstract interpretation to perform efficient extraction of a minimal automaton describing the state dynamics of a given RNN. We use Angluin's L* algorithm as a learner and the given RNN as an oracle, employing abstract interpretation of the RNN for answering equivalence queries. Our technique allows automaton-extraction from the RNN ...
[Full version]

Hardware and operating system interfaces should ideally enable to build performant systems that are also robust and secure. Alas, these goals frequently conflict, necessitating unavoidable compromises that promote one goal at the expense of another. I nevertheless contend that existing compromises---while inherently imperfect---are often suboptimal, and that improved tradeoffs can be devised using programmable software interfaces and hardware. I support this claim by presenting how programmable software interfaces can help to efficiently bridge the semantic ...
[Full version]

You are invited  to the first of series of meetings of technical workshop on various topics, on Wednesday, January 3rd, 2018, ate 18:30. The first lecture will be given by Maya Anderson from IBM Research on: How to Stop Fearing Code Changes: Version Management with git. Participation requires pre-registration and the lecture location will be defined in emailed confirmation. You are all invited!
[Full version]

We are happy to invite your to the third of series of meetings on career and job seeking which will be held at CS, on Wednesay, January 3rd, at 15:00, in room 337, CS Taub Building. In the program: 15:00 - Prof. Irad Yavneh: Opening speech 15:05 - Prof. Miri Ben-Chen: A lecture on Professional Promotion in the Academy 15:35 - Panel on "Academic Career" with: Prof. Miri Ben-Chen, Prof. Benny Kimelfeld, Prof. Eitan Yaakobi ...
[Full version]

We consider graph coloring and related problems in the distributed message-passing model. **Locally-iterative** algorithms are especially important in this setting. These are algorithms in which each vertex decides about its next color only as a function of the current colors in its 1-hop neighborhood. In STOC'93 Szegedy and Vishwanathan showed that any locally-iterative (\Delta + 1)-coloring algorithm requires Omega(\Delta \log \Delta + \log^* n) rounds, unless there exists "a very special type of coloring that ...
[Full version]

In recent years, applications increasingly adopt security primitives designed from the start with built-in side channel protection. A concrete example is Curve25519, which has been recently standardized in RFC-7748. Dealing away with obvious leakage sources such as key-dependent branches and memory accesses, RFC-7748 dictates that implementations should use a highly regular Montgomery ladder scalar-by-point multiplication, a unified, branchless double-and-add formula and a constant-time argument swap within the ladder. Moreover, as Curve25519 provides innate protection from ...
[Full version]

We introduce two new cryptographic notions in the realm of public and symmetric key encryption. * Encryption with invisible edits is an encryption scheme with two tiers of users: “privileged” and “unprivileged”. Privileged users know a key pair (pk,sk) and “unprivileged” users know a key pair (pk_e,sk_e) which is associated with an underlying edit e to be applied to messages encrypted. When an unprivileged user attempts to decrypt a ciphertext generated by a privileged user ...
[Full version]

It is well known that certain classes of signals can be effectively represented using a wavelet basis or a wavelet frame, keeping only a sparse number of coefficients. For example, Shearlet (discrete) frames are optimally sparse for cartoon like images. In this talk I will extend this approach to continuous wavelet systems, which comprise a continuum of "dictionary" elements. To overcome some of the challenges in the continuous realm, I will present an extension of ...
[Full version]

Computer Aided Geometric Design concerns development of algorithms and accompanying software, towards design of parts and structures with a high degree of numerical precision. Target application domains include design of automobiles, aircrafts and buildings, among others. In this talk we discuss two kernel level modeling primitives. First, we propose a comprehensive algorithmic framework for computation of solid sweeps. This involves computing the boundary of the swept volume generated by moving the input solid along the ...
[Full version]