The Taub Faculty of Computer Science Events and Talks

The availability of effective exact or heuristic solution methods for general Mixed-Integer Programs (MIPs) is of a paramount importance for practical applications. Unfortunately, in most practical large scale problems, a general-purpose MIP solver may prove not effective even after a clever tuning. In the present paper we investigate the use of a generic MIP solver as a black box tool to produce a sequence of reasonably good feasible solutions quickly. The procedure is based on ...
[Full version]

In this paper we study the identity testing problem of arithmetic read-once formulas (ROF) and some related models. A read-once formula is formula (a circuit whose underlying graph is a tree) in which the operations are {+, ×} and such that every input variable labels at most one leaf. We obtain the first polynomial-time deterministic identity testing algorithm that operates in the black-box setting for read-once formulas, as well as some other related models. As ...
[Full version]

Computer systems are everywhere, often controlling critical processes and containing sensitive secret information. However, the ubiquitous nature of computer systems also means that such systems often operate in hostile environments where they are subjected to various attacks by adversarial parties. Even if the system's security is theoretically proven under some set of assumptions, when faced with real-word attack scenarios, many theoretical assumptions become flaky, inaccurate and often completely incorrect. In this talk I will present ...
[Full version]

Existing resource management frameworks for large scale cloud systems leave unresolved the problematic tension between high resource utilization and job's performance predictability - respectively coveted by operators and users. In this talk, I will present recent efforts to resolve this tension through temporal planning: unlike popular scheduling and routing schemes, we propose mechanisms that plan the resource allocations into future time steps. Intuitively, such planning allows the operator to pack the cloud more densely, while ...
[Full version]

CS will hold a symposium honoring Prof. Johann Makowsky’s retirement on: Logic in Computer Science and Combinatorics On the agenda greetings by CS Dean Prof. Irad Yavneh and talks by: Moshe Y. Vardi Eldar Fischer Tomer Kotek Nadia Labai Elena Ravve Johann Makowsky More details and full program. The event will take place on Sunday, December 25 2016, between 13:00-17:00 at CS Taub 337, Technion. You are all invited!
[Full version]

For the past 40 years computer scientists generally believed that NP-complete problems are intractable. In particular, Boolean satisfiability (SAT), as a paradigmatic automated-reasoning problem, has been considered to be intractable. Over the past 20 years, however, there has been a quiet, but dramatic, revolution, and very large SAT instances are now being solved routinely as part of software and hardware design. In this talk I will review this amazing development and show how automated reasoning ...
[Full version]

With the proliferation of multi-core processors, shared-memory concurrent programming has become increasingly important. Nevertheless, despite decades of research, there are still no adequate answers to the following fundamental questions: 1. What is the right semantics for concurrent programs in higher-level languages? 2. Which reasoning principles apply to realistic shared-memory concurrency? Concerning the first question, the challenge lies in simultaneously allowing efficient implementation on modern hardware with compiler optimizations, while exposing a well-behaved programming model. Both ...
[Full version]

In this talk I will present some of the research I have been involved in concerning the specification and monitoring of timed, continuous and hybrid behaviors using formalism such as Signal Temporal Logic and Timed Regular Expressions. I will discuss the similarities and differences between properties/assertions and other ''measures'' which are used in many application domains to evaluate behaviors, as well as the difference between monitoring real systems during their execution and monitoring simulated models ...
[Full version]

We present recent developments in randomness extractors theory and applications to classical, long-standing, open problems such as Ramsey graphs constructions and privacy amplification protocols. This exciting progress heavily relies on two new pseudo-random primitives we call correlation breakers and independence-preserving mergers, which we discuss. Short bio: ========== Gil Cohen is a postdoctoral researcher at Princeton University working with Mark Braverman. He obtained his Ph.D. in 2015 from The Weizmann Institute of Science under the guidance ...
[Full version]

The study and design of concurrent data structures has mostly focused on verifying their correctness and proving general progress guarantees. In contrast, very little work has been done on analyzing the running time of these data structures, possibly due to a lack of a good model under which to perform such analysis. We take a step in this direction by presenting a concurrent relaxed counter which we show how to use to keep track of ...
[Full version]

Challenges in simulation, optimization and control of natural gas transmission systems and their coupling to power transmission systems are reviewed in this presentation describing research on the subject by the Grid Science Team at LANL. In this presentation I will describe opportunities but also challenges emerging in view of new dependencies between power and natural gas regional, national, and international systems. The availability of natural gas and the need for cleaner electric power have prompted ...
[Full version]

When transmitting information over a noisy channel, two approaches are common: assuming the channel errors are independent of the transmitted content and devising an error-correcting code, or assuming the errors are data dependent and devising a constrained-coding scheme that eliminates all offending data patterns. In this talk we analyze a middle road, which we call a semiconstrained system. In such a system we do not eliminate the error-causing sequences entirely, but rather restrict the frequency ...
[Full version]

Delivery of social notifications has become a popular component in many electronic services. While a small number of services for social notifications, e.g., Twitter, are dedicated, an increasingly large number of such services are provided as an enhancement of the main functionality. For example, the main music streaming service of Spotify is accompanied by an engine that delivers real-time notifications to each online user. These notifications report when a user's friend comes online and goes ...
[Full version]

In recent years, genetic studies have revolutionized our understanding of common diseases like diabetes and cancer. However, the analysis of genetic studies of disease poses substantial statistical and computational challenges, owing to the data collection scheme and to the extremely large data dimensionality. We propose a unified modeling framework that addresses these challenges in order to solve the three main problems associated with genetic disease studies: Searching for disease causing mutations, predicting risk of being ...
[Full version]

Facebook Engineers will come to CS and will offer you full-time positions and internship opportunities on Wednesday, December 14, 2016, 12:30-13:30, at CS Taub Building, Room , 3rd Floor, and will give an Engineering Tech Talk on: Data Driven Engineering @ Facebook: At Facebook, we use data to understand growth, as well as identify and execute on enhancements to product experiences. Given hundreds of millions of users worldwide and multiple products across different domains, the ...
[Full version]

We will start with presenting the basic notions of (co)homomology of simplical complexes (which requires only basic linear algebra over the field of order 2) and then we will indicate its relevance for several topics in computer science and combinatorics: 1) Property testing.: 2) Quantum error correcting codes : 3) High dimensional expanders.: 4) Random simplical complexes.
[Full version]

Reliable performance evaluations require representative workloads. This has led to the use of accounting logs from production systems as a source for workload data in simulations. I will survey 20 years of ups and downs in the use of workload logs, culminating with the idea of resampling with feedback. It all started with the realization that using workload logs directly suffers from various deficiencies, such as providing data about only one specific situation, and lack ...
[Full version]

Capsule Endoscopy for the Colon is an emerging market with the potential to become a screening tool for detecting colorectal cancer. The patient swallows a pill that includes cameras which provide a video of the digestive system. If a significant polyp, which is a precursor to colorectal cancer, is detected with the capsule, the patient will be referred to Colonoscopy to have it treated. In this talk the challenge of detecting polyps will be introduced. ...
[Full version]

There are *many* photos in the world. An average user might have thousands of photos in their personal photo collection. We have reached a point where photo acquisition is trivial, and the next challenge lies in arranging and easily editing such large photo collections. I will start the talk by briefly surveying a few of our works that aim to arrange large collections, and to provide fast (yet sophisticated) image manipulation techniques. Next, I will ...
[Full version]

OpenMP is an API for writing multithreaded, shared memory parallelism, it consists of a set of compiler directives none intrusive to the original serial code. In addition I shall present parfor concept in MATLAB which is useful where you need many loop iterations of a simple calculation.
[Full version]

The Internet infrastructure relies almost entirely on network protocols that are based on open standards. However, the majority of network devices on the Internet, e.g. routers and switches, are proprietary and closed source. Hence, there is no straightforward way to analyze them. Specifically, one cannot easily and systematically identify deviations of a network device's protocol implementation from the protocol's standard. Such deviations may degrade the security or resiliency of the network. In this talk I ...
[Full version]

We address a new online problem, referred to as min-cost perfect matching with delays (MPMD), where requests arrive in a continuous time online fashion at the points of a finite metric space $mathcal{M}$ and should be served by matching them to each other. The algorithm that knows $mathcal{M}$ in advance is allowed to delay its matching commitments, but this does not come for free: the total cost of the algorithm is the sum of metric ...
[Full version]

Network Function Virtualization (NFV) is a novel paradigm that enables flexible and scalable implementation of network services on cloud infrastructure, while Network Survivability is an traditional well-studied subject for maintaining network service continuity in the presence of failures. This talk will address these two important subjects through the results of two recent papers, namely "Optimal Link-Disjoint Node-“Somewhat Disjoint” Paths"[1] and "The Actual Cost of Software Switching for NFV Chaining"[2]. In [1], we investigate a crucial ...
[Full version]

Within the wide field of sparse approximation, convolutional sparse coding (CSC) has gained increasing attention in recent years. This model assumes a structured-dictionary built as a union of banded Circulant matrices. Most attention has been devoted to the practical side of CSC, proposing efficient algorithms for the pursuit problem, and identifying applications that benefit from this model. Interestingly, a systematic theoretical understanding of CSC seems to have been left aside, with the assumption that the ...
[Full version]

Network providers must dynamically allocate scarce physical resources among their clients to maximize benefit. Network pricing is one way for providers to maximize client benefit by allowing them to share available bandwidth according to their willingness to pay for it. The resulting allocation grants additional bandwidth to those who need it the most, while decreasing the bandwidth of those who need it the least. Existing queueing algorithms use the results of pricing schemes as weights ...
[Full version]

The interest in subspace codes has increased recently due to their application in error correction for random network coding. In order to study their properties and find good constructions, the notion of cyclic subspace codes was introduced by using the extension field structure of the ambient space. However, to this date there exists no general construction with a known relation between k, the dimension of the codewords, and n, the dimension of the entire space. ...
[Full version]

Distance based reconstruction methods of phylogenetic trees consist of two independent parts: first, inter-species distances are estimated assuming some stochastic model of sequence evolution; then the inferred distances are used to construct the tree. These methods rely heavily on comparisons of the inferred distances and of linear combinations thereof. In this work we show how to accurately and efficiently compare two evolutionary distances, and set the ground for more complicated comparisons (e.g., comparisons of sums ...
[Full version]

A celebrated result by Impagliazzo and Wigderson is that under complexity theoretic hardness assumptions, every randomized algorithm can be transformed into one that uses only logarithmically many bits, with polynomial slowdown. Such algorithms can then be completely derandomized, with polynomial slowdown. In the talk I will discuss recent work attempting to extend this approach to: 1. Randomized algorithms that err with probability $1-\epsilon$ for small $\epsilon$. (Here, the goal is to minimize the number of ...
[Full version]

Users of solid-state disks and mobile devices may benefit from the ability to hide sensitive data in a manner that disallows powerful adversaries from detecting that data has been hidden. In this talk I present a new technique to achieve this goal by manipulating the voltage level of randomly selected flash cells to encode two bits (rather than one), such that one bit is "public" and the other is private. Our technique leverages the inherent ...
[Full version]

Hyperspectral (HS) images or "hyperspectral data-cubes" contain radiance spectrum information at high spectral resolution for each point in the scene. Until recently, acquiring such information involved expensive, bulky equipment which required long exposure times - making HS imaging impractical for "natural" imaging (ground-level, horizontally viewed scenes). The talk will present a novel methodology, allowing high-accuracy estimation of HS information from unseen scenes using only RGB data acquired from a consumer-grade camera. This is achieved by ...
[Full version]

Projected gradient descent (PGD), and its close variants, are often considered the methods of choice for solving a large variety of machine learning optimization problems, including empirical risk minimization, statistical learning, and online learning. This is not surprising, since PGD is often optimal in a very appealing information-theoretic sense. However, for many problems PGD is infeasible both in theory and practice since each step requires to compute an orthogonal projection onto the feasible set. In ...
[Full version]

Google's Cloud & Networking group will visit CS and will give a Tech Talk presenting ithe technical highlights and challenges of building a project called "Effingo" - Google's internal (very) large scale file copy service, as well as on their hiring process. Please pre-register.
[Full version]

Many algorithms for the computation of correspondences​ ​between deformable shapes rely on some variant of​ ​nearest neighbor matching in a descriptor space. Such are,​ ​for example, various point-wise correspondence recovery​​ algorithms used as a post-processing stage in the functional​ ​correspondence framework. Such frequently used techniques​ ​implicitly make restrictive assumptions (e.g., nearisometry)​ ​on the considered shapes and in practice suffer from lack of accuracy and result in poor surjectivity. We​ ​propose an alternative recovery technique capable ...
[Full version]

Everyone wants to program with "high-level concepts", rather than meddle with the fine details of the implementation, such as pointers, network packets, and asynchronous callbacks. This is usually achieved by introducing layers of abstraction - but every layer incurs some overhead, and when they accumulate, this overhand becomes significant and sometimes prohibitive. Optimizing the program often requires to break abstractions, which leads to suboptimal design, higher maintenance costs, and subtle hard-to-trace bugs. I will present ...
[Full version]

In the classical Node-Disjoint Paths (NDP) problem, the input consists of an undirected $n$-vertex graph $G$, and a collection $M=\{(s_1,t_1),\ldots,(s_k,t_k)\}$ of pairs of its vertices, called source-destination, or demand, pairs. The goal is to route the largest possible number of the demand pairs via node-disjoint paths. The best current approximation for the problem is achieved by a simple greedy algorithm, whose approximation factor is $O(\sqrt n)$, while the best current negative result is a roughly ...
[Full version]

Please see details in the Hebrew page.
[Full version]

Finding correspondence between 3D shapes is one of the prototypical problems in computer graphics, geometric processing, and vision. Different flavors of this problem arise in applications ranging from texture mapping and animation to marker-less motion capture. The recent progress in the development of commercial real-time 3D scanning technology has brought the need for fast, accurate, and reliable correspondence methods capable of dealing with real-world noise and artifacts. Particularly hard settings of the correspondence problem include ...
[Full version]

An (n, N, k, m, t)-batch code abstracts the following distributed database problem: n items are to be distributed among m servers in such a way that any (multi)set of k items can be retrieved by reading at most t items from each server with the condition that the total storage over m servers be bounded by N. Combinatorial batch codes (CBCs) are replication based variants, i.e., for these codes each of the N stored ...
[Full version]

Learning from membership queries has flourished due to its many applications in group testing, blood testing, chemical leak testing, chemical reactions, electric shorting detection, codes, multi-access channel communications, molecular biology, VLSI testing and AIDS screening. Many of the new applications raised new models and new problems and in many of those applications the function being learned can be an arbitrary function that depends on few variables. We call this class of functions $d$-Junta, where $d$ ...
[Full version]

Branching program is a combinatorial model of computation which models space in Turing machines. One of the holy grail in complexity theory is to understand the relationship between efficient space (class L, logarithmic space in input) and efficient time (class P, polynomial time in input). It is widely believed that there are problems which are in P but not in L. To prove that a problem is not in L it is enough to prove ...
[Full version]

Network function cloudification aims at executing virtual network functions on the cloud to enable rapid deployment of new services with greater flexibility and better CAPEX and OPEX efficiencies. While the cloud provides a convenient overlay abstraction to the underlying compute and network infrastructure, the data will often travel over the network to be processed by these virtual functions and possibly travel again to be consumed elsewhere. Network overlays suppress important topological details of the underlying ...
[Full version]

​We present a general framework for solving partial differential equations on manifolds represented by meshless points, i.e. point clouds, without parameterization or connection information. Our method is based on a local approximation of the manifold as well as functions defined on the manifold, such as using least squares, simultaneously in a local intrinsic coordinate system constructed by local principal component analysis using K nearest neighbors. Once the local reconstruction is available, differential operators on the ...
[Full version]

Functional Safety and IT-Security are the most important quality factors for future real-time embedded systems. This talk will cover state of the art techniques and research results covering functional safety and IT-Security in relation to functional safety. Focus mainly on your knowledge and experience in the Automotive Industry. Prof. Dr. Jürgen Mottok has been involved more than 20 years in the development and assessment of safety critical automotive systems at SIEMENS VDO Automotive AG, now ...
[Full version]

Embedded real-time systems focusing on functional safety applications must be error-free and have to master faults occurring during operation. In this talk we will deal with types of faults and how the system architecture will control them. We will ask the question: "How much safety do we need?" In the future, the automobile will drive autonomously and we will outline the question whether the valid automotive standard ISO26262 can be adapted for this purpose. Focus ...
[Full version]

One of the most exciting possibilities opened by deep neural networks is end-to-end learning: the ability to learn tasks without the need for feature engineering or breaking down into sub-tasks. This talk will present three cases illustrating how end-to-end learning can operate in machine perception across the senses (Hearing, Vision) an​​d even for the entire perception-cognition-action cycle. The talk begins with speech recognition, showing how acoustic models can be learned end-to-end. This approach skips the ...
[Full version]

This talk will present some basic concepts to help (aspiring) users of 3D printing. The talk will focus on the pipeline and use of the common FDM printers found in the CGGC lab.
[Full version]

One of the fundamental tasks in many computational settings is finding the maximum cut. Surprisingly, it has been insufficiently studied in the classic distributed settings, where vertices communicate by synchronously sending messages to their neighbours according to the underlying graph. In the talk we will discuss the techniques we developed for approximating Max-Cut in the distributed setting, and present several approximation algorithms obtained using them.
[Full version]

We consider the Multiway Cut problem, a basic graph partitioning problem in which the goal is to find the minimum weight collection of edges disconnecting a given set of special vertices called terminals. Multiway Cut admits a well known simplex embedding relaxation, where rounding this embedding is equivalent to partitioning the simplex. Current best known solutions for the problem are comprised of a mix of several different ingredients, resulting in intricate algorithms. Moreover, the best ...
[Full version]

On-chip heterogeneity has been shown to be an effective mechanism to improve execution efficiency for general purpose and embedded computing. Existing heterogeneous designs either feature a single ISA or multiple ISAs that statically partition work. In this talk, PhD candidate from UC San Diego Ashish Venkat, will describe his research that enables programs to cross a heretofore forbidden boundary -- the ISA. In particular, Venkat will describe a compiler and runtime strategy for swift and ...
[Full version]

In the analysis of high-dimensional data, manifold learning methods are used to reduce the dimensionality of the data, while preserving local neighborhoods and revealing meaningful structures. Out-of-sample function extension techniques are then used for analyzing new points, yet these techniques are inherently limited for handling outliers. I present an analysis of these limitations and propose a new Multiscale Anomaly Detection approach that overcomes them. We have applied our approach to challenging remote sensing datasets, demonstrating ...
[Full version]

The field of distributed graph algorithms deals with solving graph problems in a network (graph) of independent agents while minimizing the amount of communication between nodes. One such fundamental problem is weighted vertex cover. We show a fast distributed (2+epsilon)-approximation algorithm using the local ratio method. Due to a known lower bound, the number of communication rounds our algorithm achieves is tight for all constant values of epsilon. This is a joint work with Keren ...
[Full version]

In a profoundly influential paper, Shannon introduced information theory and used it to study the one-way data transmission problem, initiating the study of data compression. In the last few decades, the study of efficient communication using tools from information theory has led to many interesting results. In this talk we will focus on the interactive analog of data compression, the interactive compression question: Can we compress interactive communication to its information content? The standard way ...
[Full version]

Controlled light conditions improve considerably the performance of most computer vision algorithms. Dynamic light conditions create varying spatial changes in color and intensity across the scene. These condition, caused by a moving shadow for example, force developers to create algorithms which are robust to such variations. We suggest a computational camera which produces images that are not influenced by environmental variations in light conditions. The key insight is that many years ago, similar difficulties were ...
[Full version]

An inconsistent database is a database that violates some of the integrity constraints assumed to hold. Managing data inconsistency has been one of the major challenges in the research and practice of database management. In the Big Data era, applications often collect and integrate information from a large collection of information sources, containing imprecise measures, natural-language text, and so on. Sometimes these data sources contain inconsistent and imprecise information or they disagree with one another, ...
[Full version]

With intelligent processing, cameras have great potential to link the real world and the virtual world. We review advances and opportunities for algorithms and applications that retrieve information from large databases using images as queries. For rate-constrained applications, remarkable improvements have been achieved over the course the MPEG-CDVS (Compact Descriptors for Visual Search) standardization. Beyond CDVS lie applications that query video databases with images, while others continually match video frames against image databases. Exploiting the ...
[Full version]

The SIFT method is the first efficient image comparison model. It was the first method to propose a practical scale-space sampling and to put in practice the theoretical scale invariance in scale-space theory. SIFT associates with each image a list of translation, rotation and scale invariant features used for comparison with other images. Despite countless applications and an avalanche of variants, not much has been done to really understand this central algorithm and to find ...
[Full version]

CS invites all degrees students to attend the first information fair and to check out the variety of CS projects, seminars and advanced courses to be given in the 2016-2017 academic year. The event will take place on Wednesday's Noon, 12:30-14:30, at the lobby of CS Taub Building. You are all invited!
[Full version]

Guruswami and Smith (Journal of the ACM, to appear) present a construction for an explicit codes for additive channels (channels that can induce a fixed error) that succeeds with high probability. Furthermore Guruswami and Smith gave an explicit construction against channels of size n^c if we provide the encoding and decoding with public shared randomness (Monte-Carlo construction). In this paper we solve several open problems posed by Guruswami and Smith: 1. We show a fully ...
[Full version]

Image and texture synthesis is a challenging task that has long been drawing attention in the fields of image processing, graphics, and machine learning. This problem consists of modelling the desired type of images, either through training examples or via a parametric modeling, and then generating images that belong to the same statistical origin. This work addresses the image synthesis task, focusing on two specific families of images -- handwritten digits and face images. This ...
[Full version]

Task-based programming is an emerging paradigm that simplifies parallel programming. Task-based models, however, mostly focus on expressing concurrency and, for the most part, do not reason about data synchronization. The recently proposed O-structures memory versioning model is intended to fill this gap and dynamically track data dependencies. The model extends the concept of register renaming to the entire memory and thereby eliminates false- and anti-dependencies between tasks. However, it only provides low-level primitives, which makes ...
[Full version]

We present a model in which extremely simple agents, able only to detect how many other agents are in a sector directly in front of them, without ever knowing actual bearings towards any other agent, come to an agreement about a common center of rotation and synchronize their orientation to form a regular polygon or a set of regular polygons, depending on a simple predefined parameter, the agent's field of view. Methods of controlling the ...
[Full version]

Computer science and optics are usually studied separately -- separate people, in separate departments, meet at separate conferences. This is changing. The exciting promise of technologies like virtual reality and self-driving cars demand solutions that draw from the best aspects of computer vision, computer graphics, and optics. Previously, it has proved difficult to bridge these communities. For instance, the laboratory setups in optics are often designed to image millimeter-size scenes in a vibration-free darkroom. In ...
[Full version]

A collision resistant hash function is a function that compresses the input, yet it is computationally infeasible to find two inputs on which it has the same output. Collision resistant hash functions are widely used for making digital signatures efficient and have many other applications in cryptography. The talk will discuss the question of minimizing the complexity of collision resistant hash functions, where complexity is measured in terms of algebraic degree, circuit size, or locality. ...
[Full version]

Auto-parallelization of programs that have not been developed with parallelism in mind is one of the holy grails in computer science. It requires understanding the source code's data flow to automatically distribute the data, parallelize the computations, and infer synchronizations where necessary. We will discuss our new LLVM-based research compiler Polly-ACC that enables automatic compilation to accelerator devices such as GPUs. Unfortunately, its applicability is limited to codes for which the iteration space and all ...
[Full version]

Parameter estimation is performed by fitting data measurements to a model using Bayesian statistics, assuming additional prior information. The estimation requires a numerical solution of a large scale optimization problem, whose objective traditionally includes data fidelity and regularization terms. In this talk we will concentrate on parameter estimation of physical models, obtained by solving optimization problems that are constrained by partial differential equations (PDEs). We will focus on the 3D Full Waveform Inversion, which arises ...
[Full version]

Motion capture is the process of recording the movement of objects or people by sensors and transforming it into computer readable format. To achieve best results, the environment needs to be optimally covered by the sensors. I will propose my latest advances in optimally placing and orienting multiple cameras for motion capture systems in risky industrial settings. Sample objectives for the optimization are to reconstruct an object most accurately, to maximally cover important regions of ...
[Full version]

The 5th TCE Summer School on Cyber and Computer Security will be held on Sunday-Thursday, September 4th-8th, 2016. Location: Room 1003, Meyer Building (EE), Technion, Haifa. Attendance is free, but registration is required Topics include innovation and entrepreneurship in big data: from preaching to (effective, efficient and secure) data science practices. The program includes a two days’ workshop by Intel experts, presenting all security technologies integrated into computing solution, including future technologies not yet in ...
[Full version]

Real systems for mining dynamic data streams should be able to detect changes that affect the accuracy of their model. A distributed setting is one of the main challenges in this kind of change detection. In a distributed setting, model training requires centralizing the data from all nodes (hereafter, synchronization), which is very costly in terms of communication. In order to minimize the communication, a monitoring algorithm should be executed locally at each node, while ...
[Full version]

In modern architectures memory operations may be reordered and executed non-atomically. Consistency is guarantied for single thread execution, but not across threads. Both software and hardware Relaxed Memory Models are defined to capture the possible execution behaviors. Each Relaxed Memory Model poses its own challenges to successful program verification, and, in cases where verification fails, poses its own challenges to automatic synthesis of program corrections. Our work covers Intel's x-86 TSO and PSO hardware buffered ...
[Full version]

Text categorization is a prominent task of labeling documents with predefined categories. The main approach to performing text categorization is learning from labeled examples. For many tasks, it may be difficult to find examples in one language but easy in others. The problem of learning from examples in one or more languages and classify (categorize) in another is called Cross-language learning. Existing approaches for solving this problem relies on translation of the texts from all ...
[Full version]

Studying the evolution of software is an emerging topic, receiving much attention and focus as every line of code written today is more likely to belong to an existing piece of software. Our work is an effort at modeling and quantifying the semantic impact of these changes, over the scope of a single procedure, a whole program as well as compiled binaries.
[Full version]

Generic frame fields are important for many applications in computer graphics such as texture mapping and quadrangulation. The vast majority of existing approaches rely on computing an intermediate metric induced by an ambient Euclidean space, thus increasing the complexity and restricting the space of possible metrics. Other, intrinsic approaches, rely on a rather complicated formulation and require numeric approximations. We propose an intrinsic approach that obviates the necessity of an embedding and has a simple ...
[Full version]

Collaboration between multiple robots (or agents) pursuing common or individual tasks is important in numerous problem domains, including cooperative autonomous navigation, mapping and 3D reconstruction, tracking and active sensing. A key required capability is to autonomously determine robot actions while taking into account different sources of uncertainty. The corresponding approaches, known as belief space planning, represent the states of interest (e.g. robot or camera pose, objects) via a probability distribution function and reason how the ...
[Full version]

Mobile phones and tablets are becoming the main devices to access the Internet, and wireless proximity technology has improved significantly over the last years. This creates the right environment for mobile applications that rely on proximity and close peers to improve their performance. One possible use of this kind of applications applies to Opportunistic Networking, leveraging social contacts to efficiently transfer and store information. This generates the need of understanding the patterns and social characteristics ...
[Full version]

Software Defined Networks (SDN) is a new approach to computer networking that provides simple network management by the abstraction of higher level functionality. The SDN framework enables flexible and frequent updates, and it introduces challenges in terms of consistency and scalability. This seminar considers the time-based approach to updates in SDN networks, which has recently been proposed by Mizrahi and Moses. We demonstrate a time-based approach that provides lower overhead, shorter time duration, and improved ...
[Full version]

Switch codes are a coding scheme which enables one to encode the input packets into the banks such that the packet requests by the output ports can be answered efficiently. Mathematically speaking, a switch code is required to satisfy the following property. Assume that there are n​ input ports, k​ output ports, and m​ banks. In each generation the n​ packets from the input ports are encoded into m​ packets which are stored in the ...
[Full version]

Resource allocation problems arise in a wide range of applications. In many of these classic problems, we are given a set of requests competing for resources, where each request utilizes predefined amounts of the resources. We seek a feasible allocation of the resources, subject to availability constraints, so as to maximize (or, minimize) certain objective function. However, in recent applications, such as flex-grid all-optical networks, cloud computing, and cellular networks, resource requests may be partially ...
[Full version]

It is often the case in biological measurement data that results are given as a ranked list of quantities, for example differential expression (DE) of genes as inferred from microarrays or RNA-seq. Recent years brought considerable progress in statistical tools for enrichment analysis in ranked lists. Several tools are now available that allow users to break the fixed set paradigm in assessing statistical enrichment of sets of genes. Continuing with the example, these tools identify ...
[Full version]

In this work we revisit the problem of visual odometry. Visual odometry is the process of estimating the motion of the camera by examining the changes that the motion induces on the images made by it. The approach we propose exploits a scene structure typical for that seen by a moving car and is suitable for use in either the stereo or the monocular setting. We recover the rotation and the translation separately, thus dealing ...
[Full version]

While 3D movies are gaining popularity, viewers in a 3D cinema still need to wear cumbersome glasses in order to enjoy them. Automultiscopic displays provide a better alternative to the display of 3D content, as they present multiple angular images of the same scene without the need for special eyewear. However, automultiscopic displays cannot be directly implemented in a wide cinema setting due to variants of two main problems: (i) The range of angles at ...
[Full version]

In recent years, due to the spread of multi-level nonvolatile memories (NVM), q-ary write-once memories (WOM) codes have been extensively studied. By using WOM codes, it is possible to rewrite NVMs t times before erasing the cells. The use of WOM codes enables to improve the performance of the storage device, however, it may also increase errors caused by inter-cell interference (ICI). In this talk I will present WOM codes that restrict the imbalance between ...
[Full version]

Over the past 30 years numerous algorithms have been designed for symmetry breaking problems in the LOCAL model, such as maximal matching, MIS, vertex coloring, and edge-coloring. For most problems the best randomized algorithm is at least exponentially faster than the best deterministic algorithm. In this paper we prove that these exponential gaps are necessary and establish connections between the deterministic and randomized complexities in the LOCAL model. Each result has a very compelling take-away ...
[Full version]

With the rapid development of parallel platforms, optimistic methods become increasingly common. However, existing memory management schemes for such algorithms imposes high performance overhead and are considered notoriously complex. In this talk, I will describe how memory management can benefit from optimistic methods. I will start by presenting a memory reclamation scheme for lock-free data structures, which operates by turning all reads into optimistic ones. This resulting lock-free scheme is easy to incorporate, imposes low ...
[Full version]

We are interested in public-setup Zero-Knowledge proof systems for natural languages in NP with poly-logarithmic Verifier and quasi-linear Prover. Systems having such properties can be compiled efficiently to succinct publicly verifiable arguments, that could in practice effect our lives in many aspects. Although the theory behind such systems has been studied for about 3 decades, practical solutions do not yet achieve desired requirements. In a "Theory to Practice and back again" research we have made ...
[Full version]

Decisions taken by agents in distributed and multi-agent systems depend on their local information. A novel formulation of the connection between knowledge and action in such systems allows new insights into the design of correct and efficient protocols. This talk will discuss and illustrate how a theory of what agents know about the world and about the knowledge of others can help in the design and analysis of distributed protocols for well-known problems. The talk ...
[Full version]

We introduce a framework allowing domain experts to manipulate computational terms in the interest of deriving better, more efficient implementations. It employs deductive reasoning to generate provably correct efficient implementations from a very high-level specification of an algorithm, and inductive constraint-based synthesis to improve automation. Semantic information is encoded into program terms through the use of refinement types. In this paper, we develop the technique in the context of a system called Bellmania that uses ...
[Full version]

Virtualization is the ability of modern computer systems to run guest Virtual Machines (VMs). The VM host exposes various I/O devices to its guests such as the Network Interface Controller (NIC), hard disk, etc. Para-virtual I/O is a common technique for presenting the guest VM with an interface similar, but not identical, to the underlying hardware. Such interfaces are called virtual I/O devices, and their behavior is emulated by the VM host. This emulation must ...
[Full version]

CS students in the course "Project in Advanced Programming" invite you to an Arduino project fair on Tuesday, June 21, 2016, between 12:3.30, in the lobby of CS Taub Building. The fair will exhibit new Arduino and Android projects as well as yearly projects. You are all invited! Here are the presenting projects: Arduino Bothhoven: Songs are stored on the cloud, streamed to the robotic pianist, and help you enjoy your time. Cipher: Cipher is ...
[Full version]

In this paper we describe a new method for detecting and counting a repeating object in an image. While the method relies on a fairly sophisticated deformable part model, unlike existing techniques it estimates the model parameters in an unsupervised fashion thus alleviating the need for a user-annotated training data and avoiding the associated specificity. This automatic fitting process is carried out by exploiting the recurrence of small image patches associated with the repeating object ...
[Full version]

Write-once memory (WOM) is a storage device consisting of q-ary cells that can only increase their values. A WOM code is a scheme to write messages to the memory without decreasing the cells' levels. There are four models of WOM which depend on whether the encoder and decoder are informed or uninformed with the previous state of the memory. The WOM capacity of the four models was extensively studied by Wolf et al. for the ...
[Full version]

Cassandra is a highly popular distributed data store. It supports flexible consistency semantics over a wide-column data access model. In our work, we propose protocols for hardening Cassandra against Byzantine failures, in which some nodes in the system may act arbitrarily, including in a malicious manner. We examine several alternative design choices and compare between them both qualitatively and empirically by using the Yahoo! Cloud Serving Benchmark (YCSB) performance benchmark. We report on our findings ...
[Full version]

A tangent vector field on a surface is the generator of a smooth family of maps from the surface to itself, known as the flow. Given a scalar function on the surface, it can be transported, or advected, by composing it with a vector field's flow. Such transport is exhibited by many physical phenomena, e.g., in fluid dynamics. In this paper, we are interested in the inverse problem: given source and target functions, compute a ...
[Full version]

Contention resolution schemes are an algorithmic technique originally devised in the context of submodular functions maximization. Recent works, however, have extended contention resolution schemes into a robust tool with many applications in various fields such as stochastic optimization, auctions and prophet inequalities. In this talk I will present the concept of contention resolution schemes, and will give a few examples demonstrating some of the power of this technique. The talk is based on the papers: ...
[Full version]

Research on the application of game theory in the context of networking has focused on non-cooperative games, where the selfish agents cannot reach a binding agreement on the way they would share the infrastructure. Many approaches have been proposed for mitigating the typically inefficient operating points. However, in a growing number of networking scenarios selfish agents are able to communicate and reach an agreement. Hence, the degradation of performance should be considered at an operating ...
[Full version]

Gilad Barkai from Google's AUTO-COMPLETE team will give a talk about the system - its daily use, challenges and different features as the programmer sees them. Please pre-register.
[Full version]

Modern discrete GPUs have been the processors of choice primarily for compute-intensive applications, but using them in large-scale data processing is extremely challenging. Unfortunately, they do not provide important I/O abstractions that have long been established in CPU context, like memory mapped files, which shield programmers from the complexity of buffer and I/O device management. However, implementing these abstractions on GPUs poses a problem: the limited GPU virtual memory hardware does not support page faults ...
[Full version]

​Computer Science (CS) is the new language of science as Mathematics was until the Twentieth Century. CS generalizes mathematics toward being more versatile. In this talk we will focus on some problems in bio-medical imaging science and discuss and how we try to address them.​ ​Tomography is the technology of combining data from multiple views (say, 2D camera snapshots) to create a comprehensive image (say, in 3D). In medicine tomography is used to gather knowledge ...
[Full version]

This work extends a recently proposed robust computational framework for constructing the boundary representation (B-rep) of the volume swept by a given smooth solid moving along a one parameter family h of rigid motions. Our extension allows the input solid to have sharp features, and thus it is a significant and useful generalization of that work. This naturally requires a precise description of the geometry of the surface generated by the sweep of a sharp ...
[Full version]

This work extends a recently proposed robust computational framework for constructing the boundary representation (B-rep) of the volume swept by a given smooth solid moving along a one parameter family h of rigid motions. Our extension allows the input solid to have sharp features, and thus it is a significant and useful generalization of that work. This naturally requires a precise description of the geometry of the surface generated by the sweep of a sharp ...
[Full version]

Irregular algorithms, such as sorting and graph algorithms, are prevalent in computer science applications. Until recently, it was assumed that the irregular algorithms also greatly benefited from the the Out-of-Order Execution engines found in most modern day CPU processors. Out-of-Order execution, using branch predictors, speculates which branch in the execution-flow should be taken and starts that execution prematurely - allowing for improved performance. This approach typically works great for regular algorithms such as matrix multiplication ...
[Full version]

On Image Processing, Visual Information Models and ... why is it so hard for us in Numeric Analysis?
[Full version]

Proof complexity studies the complexity of mathematical proofs, with the aim of exhibiting (true) statements whose proofs are always necessarily long. One well-known proof system is Hilbert’s Nullstellensatz, which shows that if the family F={f1,…,fm} of n-variate polynomials have no common solution to the system f1=…=fm=0, then there is a proof of this fact of the following form: there are polynomials G={g1,…,gm} such that f1.g1+…+fm.gm=1 is an identity. From the perspective of computer science, it ...
[Full version]

Planar shape interpolation is a classic problem in computer graphics. We present a novel shape interpolation method that blends C∞ planar harmonic mappings represented in closed-form. The intermediate mappings in the blending are guaranteed to be locally injective C∞ harmonic mappings, with conformal and isometric distortion bounded by that of the input mappings. The key to the success of our method is the fact that the blended differentials of our interpolated mapping have a simple ...
[Full version]

One of the primary challenges in single particle reconstruction in cryo-electron microscopy is to find the three-dimensional structure of a molecule using its two-dimensional noisy projection-images. We suggest a method to estimate the unknown imaging orientations of all projection-images that is independent of their distribution. Since the relative orientation between each pair of images may only be estimated up to a two-way handedness ambiguity, we suggest an efficient procedure to remove these ambiguities by casting ...
[Full version]

Conformal maps are especially useful in geometry processing for computing shape preserving deformations, image warping and manipulating harmonic functions. The Cauchy-Green coordinates are complex-valued barycentric coordinates, which can be used to parameterize a space of conformal maps from a planar domain bounded by a simple polygon. In this work, we use the Cauchy-Green coordinates to simulate 2D potential flow with interactive control, and to construct conformal maps between planar domains. The Hele-Shaw flow describes the ...
[Full version]

Conformal maps are especially useful in geometry processing for computing shape preserving deformations, image warping and manipulating harmonic functions. The Cauchy-Green coordinates are complex-valued barycentric coordinates, which can be used to parameterize a space of conformal maps from a planar domain bounded by a simple polygon. In this work, we use the Cauchy-Green coordinates to simulate 2D potential flow with interactive control, and to construct conformal maps between planar domains. The Hele-Shaw flow describes the ...
[Full version]

A major challenge in optimizing network utilization is how to reduce bottleneck links, while taking into account that bandwidth demands may be increased over time. Traffic engineering algorithms attempt to address this challenge by employing smart routing, however current solutions optimization criterion is biased towards a non-realistic setting in which demands are increased uniformly. In this talk, I will present a novel traffic engineering algorithm which is the first to consider the more realistic assumption, ...
[Full version]

In distributed systems, communication between the participants in the computation is usually the most expensive part of the computation. Theoretical models of distributed systems usually reflect this by neglecting the cost of local computation, and charging only for messages sent between the participants; in particular, we usually assume that the computation proceeds in rounds, and in each round each participant can send only a limited number of bits. We are interested in characterizing the number ...
[Full version]

produce design developed in through the years in many fronts and requirements. Today SOC products require aggressive targets in all fronts: frequency, performance, low cost, low power…… To address those new and growing requirement there needs to be a basic change in the design concept and solutions. Working in parallel convergence, parallel fronts and analyze tradeoffs together is the approach used to being the best SOC products to the market. This talk is describing the ...
[Full version]

The Sixth CS Research Day for graduate studies will be held on Tuesday, May 31, 2016, between 14:30-16:30, at the lobby of the CS Taub Building. Research Day events are opportunity for our graduate students to expose their researches using posters and presentations to CS faculty and all degrees students, Technion distinguished representatives and to high-ranking delegates from the hi-tech leading industry companies in Israel and abroad. The participating researches will be on various topics: ...
[Full version]

Both 3D models and 2D images contain a wealth of information about everyday objects in our environment. However, it is difficult to semantically link together these two media forms, even when they feature identical or very similar objects. Real-world images are naturally variable in a number of characteristics such as viewpoint, lighting, background elements, and occlusions. This variability makes it challenging to match images with each other, or with 3D shapes. We propose a joint ...
[Full version]

As a new advancement of traditional finite element method, isogeometric analysis (IGA) was proposed to integrate design and analysis. In this talk, I will present our latest research on volumetric T-spline parameterization for IGA applications. For arbitrary-topology objects, we first build a polycube whose topology is equivalent to the input geometry and it serves as the parametric domain for the following trivariate T-spline construction. Boolean operations, geometry skeleton and centroidal Voronoi tessellation based surface segmentation ...
[Full version]

A functional, computational and mechanistic, explanation of the relationship among visual attention, interpretation of visual stimuli, and eye movements, and how these produce visual behavior, seems elusive. Here, we focus on one component, how selection is accomplished for the next fixation. The popularity of saliency map models drives the inference that this is solved; but we argue otherwise. We provide arguments that a cluster of complementary, conspicuity representations drive selection, modulated by task goals and ...
[Full version]

Generalization, informally, is the ability of a learner to reflect not just its training data, but properties of the underlying distribution from which the data are drawn. When paired with empirical risk minimization, it is a fundamental goal of learning. Typically, we say that a learning algorithm generalizes if, given access to some training set drawn i.i.d. from an underlying data distribution, it returns a hypothesis whose empirical error (on the training data) is close ...
[Full version]

A significant proportion of computer system resources are devoted to applications that can inherently tolerate inaccuracies in their data, execution and communication. Hence, “approximate computing” is promising for performance and energy efficiency. However, taking advantage of approximate computing needs: language support to specify where and how to apply approximation; analysis and mechanisms that ensure good output quality; and hardware/system support that take advantage of approximation. In this talk I will describe our effort on co-designing ...
[Full version]

The 6th annual international TCE conference (Tuesday-Wednesday, May 24-25, 2016) will focus on recent trends in 3D Visual Computing: Graphics, Geometry and everything in between. Conference Chairs: Mirela Ben-Chen (CS Technion), Yoav Schechner (EE Technion) With the advent of widespread consumer devices generating 3D data, as well as the recent advances in 3D printing, computing with 3D data has become a prominent ingredient in many fields in science and engineering. Having sound theoretical foundations for ...
[Full version]

[Note the change of time and date] Generalization and simplification are deeply related to each other: simpler explanations often reveal principles that apply more generally, and predictive scientific theories often boil down to a few simple (and deep) laws. This talk is about a manifestation of the "generalization--simplification" link within computational learning theory. We will consider the following formal frameworks for `generalization' and `simplification'. (i) Probably Approximately Correct (PAC) learning -- a statistical framework that ...
[Full version]

Computing shortest paths is one of the fundamental problems of graph algorithms. The goal of *dynamic* single source shortest paths (SSSP) is to maintain a shortest path tree from a fixed source s as the edges of the graph change over time. The most general case is the fully dynamic one, where each adversarial update inserts or deletes an arbitrary edge. The trivial algorithm is to recompute SSSP after every update in O(m) time. For ...
[Full version]

The emerging trend of Software Defined Networks (SDN) has been enthusiastically explored by researchers over the last decade, as it provides the flexibility and agility to update networks dynamically, in a way that efficiently utilizes the network resources. In this talk we explore the use of accurate time and synchronized clocks as a tool for coordinating network updates in SDNs. We discuss key use cases in which using time introduces a significant advantage compared to ...
[Full version]

We present a novel framework for high-throughput live cell lineage analysis in time-lapse microscopy images. Our algorithm ties together two fundamental aspects of live cell lineage construction, namely cell segmentation and tracking, via a Bayesian inference of dynamic models. The proposed contribution exploits the Kalman inference problem by estimating the time-wise cell shape uncertainty in addition to cell trajectory. These inferred cell properties are combined with the observed image measurements within a fast marching algorithm, ...
[Full version]

Many image restoration algorithms in recent years are based on patch-processing. The core idea is to decompose the target image into fully overlapping patches, restore each of them separately, and then merge the results somehow. This concept has been demonstrated to be highly effective, leading often times to state-of-the-art results in denoising, deblurring, super-resolution, and other applications. Several questions arise from this line of work: 1) How can the local model be enforced on patches ...
[Full version]

The interest in subspace codes has increased lately due to their application in error correction for random network coding. A subspace code is a collection of subspaces of a vector space under the subspace distance $d_s(U,V)=\dim U + \dim V-2\dim(U\cap V)$. In this dissertation we discuss both purely mathematical and practical questions which involve subspace codes. The theoretical part of this work concerns a few special cases of subspace codes. Equidistant subspace codes, in which ...
[Full version]

The Anscombe transform offers an approximate conversion of a Poisson random variable into unit-variance Gaussian one. This transform is important and appealing, as it is easy to compute, and becomes handy in various inverse problems with Poisson noise contamination. Solution to such problems can be done by first applying the Anscombe transform, then applying a Gaussian-noise-oriented restoration algorithm of choice, and finally applying an inverse Anscombe transform. The appeal in this approach is due to ...
[Full version]

Gabidulin codes, originally defined over finite fields, are an important class of rank metric codes with various applications. Recently, their definition was generalized to certain fields of characteristic zero. This generalization led to new applications, such as novel constructions of space-time codes and methods for low-rank matrix recovery. The talk provides an overview of the definition, decoding and applications of Gabidulin codes over fields of characteristic zero.
[Full version]

You are invited to the final stage of the 2015-16 Amdocs Best Project Contest. The competing teams will present and talk about their projects. The event will take place on Sunday,May 8, 2015, 14:30-16:30, in Auditorium 2, CS Taub Building. You are all invited to cheer, support and be exposed to outstanding projects.
[Full version]

Motivated by the problem of zero-error communication over a broadcast channel, we introduce a new notion of graph capacity that generalizes the Shannon capacity of a graph. We derive upper and lower bounds on this quantity for arbitrary graphs, and provide a tighter upper bound for regular graphs.Joint work with Sihuang Hu.
[Full version]

The problem of computing shape correspondence is a fundamental task in Computer Graphics. While recent methods produce semantically meaningful maps globally, the maps are distorted locally. The harmonic map heat flow is a tool for local minimization of the Dirichlet energy, that measures the smoothness of a map. While the discrete harmonic map flow has been used, for example, to conformally map shapes to a sphere, when the target shape is arbitrary applying this flow ...
[Full version]

The ubiquitous nature of computer systems means that they often operate in hostile environments where they are subjected to various attacks by adversarial parties. The purpose of such attacks varies, ranging from simply corrupting the system's behaviour to a complete extraction of otherwise-unavailable secret information. My research explores theoretical and practical aspects of designing computer systems which can securely operate in hostile environments. In the talk I will describe two aspects of my research: 1. ...
[Full version]

Computational vision, visual computing and biomedical image analysis have made tremendous progress of the past decade. This is mostly due the development of efficient learning and inference algorithms which allow better and richer modeling of visual perception tasks. Hyper-Graph representations are among the most prominent tools to address such perception through the casting of perception as a graph optimization problem. In this talk, we briefly introduce the interest of such representations, discuss their strength and ...
[Full version]

Matching keypoints by minimizing the Euclidean distance between their SIFT descriptors is an effective and extremely popular technique. Using the ratio between distances, as suggested by Lowe, is even more effective and leads to excellent matching accuracy. Probabilistic approaches that model the distribution of the distances were found effective as well. This work focuses, for the first time, on analyzing Lowe's ratio criterion using a probabilistic approach. We provide two alternative interpretations of this criterion, ...
[Full version]

Due to the lack of computational power, many users don't perform the computation locally, rather outsource the computation to a remote server. This raises the problem of computational integrity. Using Probabilistically Checkable Proofs (PCP) the remote server can prove to the user (with high probability) that the computation was executed correctly. Although every language in NEXP is known to have a PCP system, previous works have not specified the process of converting instances of the ...
[Full version]

We consider the design of private prediction markets, financial markets designed to elicit predictions about uncertain events without revealing too much information about market participants' actions or beliefs. Our goal is to design market mechanisms in which participants' trades or wagers influence the market's behavior in a way that leads to accurate predictions, yet no single participant has too much influence over what others are able to observe. We study the possibilities and limitations of ...
[Full version]

In 2014, 12.4% of the population in Singapore were above 65 years of age and this is projected to increase to 19% by 2030. Among them, those living alone is likely to increase to 83,000 by 2030, up from 35,000 today. The ability to “age in place” - living where you have lived for years with reasonable quality of life, is especially important for the latter group. While Internet of Things (IoT)-enabled ambient intelligence environments ...
[Full version]

In recent years much work has been devoted to the development of image processing algorithms using local patches. The main idea in this line of work is to impose a statistical prior on the patches of the desired image. An algorithm following this path extracts all possible patches with overlaps from the image, and operates on each separately. The more advanced algorithms exploit also interrelations between different patches in the reconstruction process. In this work ...
[Full version]

We will discuss two related works:1) "Point Registration via Efficient Convex Relaxation" Point cloud registration is a fundamental task in computer graphics, and more specifically, in rigid and non-rigid shape matching. The rigid shape matching problem can be formulated as the problem of simultaneously aligning and labeling two point clouds in 3D so that they are as similar as possible. We name this problem the Procrustes matching (PM) problem. The non-rigid shape matching problem can ...
[Full version]

We present a randomized algorithm that computes a Minimum Spanning Tree (MST) in O(log^* n) rounds, with high probability, in the Congested Clique model of distributed computing. In this model, the input is a graph on n nodes, initially each node knows only its incident edges, and per round each two nodes can exchange O(log n) bits. Our key technical novelty is an O(log^* n) Graph Connectivity algorithm, the heart of which is a (recursive) ...
[Full version]

This talk considers vector network coding based on rank-metric codes and subspace codes. Our main result is that vector network coding can significantly reduce the required field size compared to scalar linear network coding in the same multicast network. The achieved gap between the field size of scalar and vector network coding is in the order of $q^{(\ell-1)t^2/\ell}-q^t$, for any $q \geq 2$, where $t$ denotes the dimension of the vector solution, and the number ...
[Full version]

CS invites you to a STARTUP DAY and recruitment by the presenting firms: Augury, Bidalgo, Checkmarx, Jelly Button, Kaltura, Outbrain, Zerto and more. In addition, lectures will be given by the firms representatives and entrepreneurs . The event will take place on Wednesday, April 20, 2016 between 12:30-14:30 at the CS Taub Lobby. The lectures will be held in Auditorium 2.. More details in the attached documents. You are all welcome!
[Full version]

Quantum money is a quantum state that can be easily verified but is hard to copy. I will start by reviewing Aaronson & Christiano's quantum money scheme. Then we'll extend their scheme to construct a tokenized blind signature scheme. A (standard) digital signature scheme uses two keys - a public key and a secret key. The secret key is used for signing an unbounded number of messages, and the public key is used to verify ...
[Full version]

Many medical images are produced daily. Compression thus plays a major role in storage and transmission applications, while enhancement, such as image de-noising, is important for diagnosis purposes. Requirements could often be conflicting, however. In order to mitigate this limitation we suggest to decompose the images into several components, where each component is distinctly described by an appropriate model and accordingly handled. For the case of ultrasound imaging, for example, we propose to describe the ...
[Full version]

Although chirp-coded signals can theoretically improve the performance of ultrasound medical imaging, they are presently of limited use. Their main advantage is better power delivery without loss of axial resolution. Their drawback, however, is that the biological tissue distorts the ultrasound echo of the chirp, and thus the shape of the received ultrasound pulse is unknown. This causes degradation in the quality of the reconstructed ultrasound image when a matched filter is used. To overcome ...
[Full version]

Rated by Ron Graham as a top-10 mathematical card trick of the 20th century, Diaconis’ mind reader is a magic trick that involves the interaction with five collaborative volunteers. Inspired by this magic, we perform in this talk a card trick that can tolerate bluffing volunteers. The theory behind this trick will be used to develop fundamental limits as well as code constructions for faster delay estimation in positioning systems.
[Full version]

We study a model of secure computation in which deviating parties are forced to pay monetary penalties, without relying on a trusted party to provide an ideal bank functionality. Our protocols work in a hybrid model where parties have access to claim-or-refund or multilock ideal functionalities, which can be efficiently realized in (a variant of) Bitcoin. In this talk we present three results: 1. Compilation of any secure multiparty computation protocol into a protocol that ...
[Full version]

In this talk, we will consider coding schemes for multi-party interactive communication over synchronous networks that suffer from stochastic noise, where each bit is independently flipped with probability ε. We analyze the minimal overhead that must be added by the coding scheme in order to succeed in performing the computation despite the noise. Our main result is a lower bound on the communication of any noise-resilient protocol over a star network with n-parties. Specifically, we ...
[Full version]

As economic systems "move" to the Internet, they can become much more complex and this new complexity often becomes their defining characteristic. We will consider a very simple scenario of this form: a single seller that is selling multiple items to a single buyer. We will discuss the question of how *complex* must the pricing scheme be in order for the seller to maximize (approximately, at least) his revenue. Based on joint works with Sergiu ...
[Full version]

To recover the three dimensional (3D) volumetric distribution of matter in an object, images of the object are captured from multiple directions and locations. Using these images, tomographic computations seek the distribution. In highly scattering media and constrained irradiance, tomography must explicitly account for off-axis scattering. Furthermore, the tomographic model and recovery must function when imaging is done in-situ, as occurs in medical imaging and ground-based atmospheric sensing. We formulate tomography that handles arbitrary orders ...
[Full version]

Multiport switches are commonly used as data processing and routing devices in computer networks. A network switch routes data packets between its multiple input and output ports. Packets from input ports are stored upon arrival in a switch fabric comprising multiple memory banks. This can lead to memory contention when distinct output ports request packets from the same memory bank, resulting in a degraded switching bandwidth. To solve this problem, switch codes are introduced by ...
[Full version]

In anticipation of 3D printers reaching millions of users, I am investigating how to allow future users to interact with the new hardware. I present a series of interactive software+hardware systems that I created to answer this question. They are characterized by two main properties. First, they produce physical output quickly, allowing users not only to see their results, but also to touch and test their mechanical properties as users work towards a solution. Second, ...
[Full version]

You are invited to the event "Leading in Technology": "Presidents on the Edge". Special guests will be Prof. Rivka Carmi, President of Ben-Gurion University and Prof. Peretz Lavie, Technion President. Einav Galili will conduct a panel on "Initiatives - Leading in Technology" hosting entrepreneurs and leaders in industry and the academy. Eexcellent students (girls) will receive scholarships by EMC and Intel. The event will take place on Wednesday, April 6, 2016 at 16:30, in Auditorium ...
[Full version]

In the past few decades, probabilistic checking techniques were shown to yield dramatic efficiency improvements in verifying proofs and approximating distance from error-correcting codes. We show connections between cryptography and ``zero-knowledge'' variants of these techniques, such as Probabilistically Checkable Proofs (PCPs) and PCPs of Proximity (PCPPs), and use them to improve the complexity of cryptographic protocols. In this talk we will present three results: 1. PCPs for NP with zero-knowledge guarantees, that can be verified ...
[Full version]

Network Function Virtualization (NFV) is a new networking paradigm where network functions are executed on commodity servers located in small cloud nodes distributed across the network, and where software defined mechanisms are used to control the network flows. This paradigm is a major turning point in the evolution of networking, and it introduces high expectations for enhanced economical network services, as well as major technical challenges. In this talk we address one of the main ...
[Full version]

The 2016 open day invite outstanding undergraduates from all universities to come to the Technion and learn about the faculties of Computer Science and  Electrical Engineering, meet faculty and graduate students and hear a fascinating talk by Dr. Yoelle Maarek, VP Research for EMEA at Yahoo, on graduate degrees from the perspective of the high-tech industry. The event will be held on Wednesday, April 6 2016, between 10:30-16:00, in CS Taub Building, room 337 (3rd ...
[Full version]

You are invited to a workshop and inauguration event of the Technion Cyber Security Research Center, in the presence of Prof. Peretz Lavie, Technion President and Dr. Evyatar Matanya, head of the National Cyber Bureau​​. The event will be held on Tuesday, April 5th, 2016, between 13:30-18:30 in Auditorium 2, CS Taub at the Technion. Attendance is free but requests pre-registration. More details and program. On the Technion Cyber Security Research Center
[Full version]

Images of outdoor scenes are often degraded by haze, fog and other scattering phenomena. In this work we show how such images can be dehazed using internal patch recurrence. Small image patches tend to repeat abundantly inside a natural image, both within the same scale, as well as across different scales. This behavior has been used as a strong prior for image denoising, super-resolution, image completion and more. Nevertheless, this strong recurrence property significantly diminishes ...
[Full version]

This work presents a novel approach for automatically repairing a program with respect to a given set of assertions. Programs are repaired using a predefined set of mutations. We impose no assumptions on the number of erroneous locations in the program, yet we are able to guarantee soundness and completeness. That is, we assure that every returned program is correct and every correct program is returned. We refer to a bounded notion of correctness, even ...
[Full version]

You are invited to a Women and Students in Technology Meeting by Microsoft. The event is held to mark The International Women's Day and will be conducted by executives and senior position representatives, including a professional panel on Career Management. More details and pre-registration.
[Full version]

One of the major challenges of the research in circuit complexity is proving super-polynomial lower bounds for de-Morgan formulas. Karchmer, Raz, and Wigderson suggested to approach this problem by proving that formula complexity behaves “as expected” with respect to the composition of functions. They showed that this conjecture, if proved, would imply super-polynomial formula lower bounds. .. In this talk, I will present the background on this conjecture and the known results. I will then ...
[Full version]

In this talk, I introduce a frequency based cache admission policy in order to boost the effectiveness of caches subject to skewed access distributions. Given a newly accessed item and an eviction candidate from the cache, our scheme decides, based on the recent access history, whether it is worth admitting the new item into the cache at the expense of the eviction candidate. Realizing this concept is enabled through a novel approximate LFU structure called ...
[Full version]

e-NABLE is an online community that designs, customizes, fabricates, and disseminates 3D-printed prosthetic hands and arms for free. E-NABLE extends the methods and philosophies of Open Source software to hardware, assistive technology development, and human resources. I'll tell the story of this remarkable community, and discuss emerging trends and opportunities for academic research and trans-academic collaboration. Bio: A biological psychologist, inventor, entrepreneur, human-computer interaction researcher, and digital community organizer, Jon Schull is the creator of ...
[Full version]

In this talk I will first give a brief overview of the topic of bi-level mathematical programming, in which the solution of an inner optimization problem is used within the objective function of a outer problem. These problems were originally studied in the context of two-player games but have been recently applied to computer vision and machine learning applications. I will then discuss current work in our group that applies bi-level optimization to the problem ...
[Full version]

We propose a framework which infers mid-level visual properties of an image by learning about ordinal relationships. Instead of estimating metric quantities directly, the system proposes ordinal relationship estimates for pairs of points in the input image. These probabilistic ordinal measurements are then aggregated and globalized to create a full output map of continuous metric measurements. Estimating order relationships between pairs of points has several advantages over metric estimation: it requires solving a simpler problem ...
[Full version]

TCE in Collaboration with Caesarea Rothschild Institute will hold a Lightweight Crypto Day. Lightweight cryptography is an emerging field of research trying to address low-end devices that need security such as RFID tags or devices connected to Internet of Things (IoT). As known security solutions cannot be accommodated in these restricted environments, lightweight cryptography has to offer meaningful security at the same time as reducing the cost of implementation to its bare essentials: footprint-wise (low ...
[Full version]

You are invited to a seminar on "Network Virtualization" held by Amdocs on Sunday, March 27 at 16:30, in class 3, CS Taub Building. Detailed agenda in the attached poster and pre-registration.
[Full version]

Non-binary low-density parity-check (LDPC) codes significantly outperform their binary counterparts. Moreover, non-binary LDPC codes are especially good for the channels with burst errors and high-order modulations. Unfortunately, their decoding complexity is still large, that is why iterative hard and soft-reliability based decoding majority algorithms are of considerable interest for high-throughput practical applications. We investigate the error-correcting capabilities of non-binary LDPC codes decoded with a hard-decision low-complexity majority algorithm, which is a generalization of the bit-flipping ...
[Full version]

The field of distributed graph algorithms deals with solving graph problems in a network (graph) of independent agents while minimizing the amount of communication between nodes. One such fundamental problem is weighted vertex cover. We show a fast distributed (2+epsilon)-approximation algorithm using the local ratio method. Due to a known lower bound, the number of communication rounds our algorithm achieves is tight for all constant values of epsilon. This is a joint work with Keren ...
[Full version]

The high speed I/O market in client platforms is going through major changes in the last few years, with latest silicon process that enable high bit rate on the wire while using low power to drive those bits. Where are those trends leading ? what is the impact on the user ? In this lecture we will review the changes in the USB & Thunderbolt technologies over the last few years, and will look forward ...
[Full version]

Geometrical understanding of bendable and stretchable structures is crucial for many applications where comparison, inference and reconstruction play an important role. Moreover, it is the first step in quantifying normal and abnormal phenomena in non-rigid domains. Moving from Euclidean (straight) distances towards intrinsic (geodesic) measures, revolutionized the way we handle bendable structures, but did not take stretching into account. Human organs, such as the heart, lungs and kidneys, are great examples for such models. In ...
[Full version]

Reed-Muller codes encode an m-variate polynomial of degree r by evaluating it on all points in {0,1}^m. Its distance is 2^{m-r} and so it cannot correct more than that many errors/erasures in the worst case. For random errors one may hope for a better result. In his seminal paper Shannon exactly determined the amount of errors and erasures one can hope to correct for codes of a given rate. Codes that achieve Shannon’s bound are ...
[Full version]

The Department of Computer Science and WomenTechPro invite you and you open the annual Spring Hakton HackPRO! where? Computer Science Department, Technion. when? Wednesday-Thursday-Friday in the first week of the spring semester, 16-18 March 2016 Think wearable technology! The Intel Edison is a tiny computer with impressive processing capability and a suitable platform for wearable technology. Crew representatives will be able to participate in the workshop Adisonim and the team that will make it run ...
[Full version]

Submodular maximization captures both classical problems in combinatorial optimization and recent more practical applications that arise in other disciplines, e.g., machine learning and data mining. Typically, the size of the inputs in these applications is very large. Hence, it is interesting to devise algorithms that in addition to providing a provable guarantee on the quality of the output are also very fast and simple. In this talk we will focus on maximization of a submodular ...
[Full version]

Good engineering practice uses the characteristics of existing technologies to optimize implementation. Often, this will mean that design techniques optimal in a previous generation prove impractical or even unusable when a new technology becomes dominant. This rule is all too often forgotten, which we will demonstrate in two problems of computer design: Field-Programmable Gate Arrays (FPGA) and hardware prefetchers (providing the ability to fetch data early in anticipation of the need). FPGAs are extremely useful ...
[Full version]

The objective of low-delay codes is to protect communication streams from erasure bursts by minimizing the time between the packet erasure and its reconstruction. Previous work has concentrated on the constant-delay scenario, where all erased packets need to exhibit the same decoding delay. We consider the case of heterogeneous delay, where the principal objective is to minimize the average delay across the erased packets in a burst. This new model is motivated by communications in ...
[Full version]

The main topic of our research is creating a multi-scale dictionary learning paradigm for sparse and redundant signal and image representations. The appeal of a multi-scale dictionary is obvious - in many cases data naturally comes at different scales. To date, popular dictionary based approaches are limited to a single scale, and small signals/patch sizes. Multi-scale approaches, on the other hand, are typically analytic, with little to no adaptation to the data. A multi-scale dictionary ...
[Full version]

In this work, we consider the list update problem as defined in the seminal work on competitive analysis by [Sleator, Tarjan 85]. In this problem, a sequence of requests, consisting of items to access in a linked list, is given. After an item is accessed it can be moved to any position forward in the list at no cost (free exchange), and, at any time, any two adjacent items can be swapped at a cost ...
[Full version]

Parkinson’s Disease (PD) is a progressive, degenerative disorder of the central nervous system. It is characterized by significant motor symptom, such as: tremor, slowness of movement, and gait deficiencies; but also entails non-motor symptoms, such as: depression, cognitive slowness and sleep difficulties. Intel Corporation is running a joint project with the Michael J. Fox foundation for Parkinson’s research, to promote research on PD, as well as patients’ daily care. As part of this project, sensory ...
[Full version]

In this work, we analyze privacy in a distributed setting, where privacy is measured in an information theoretic way (i.e. no cryptographic assumptions). The three complexity measures we analyze are (1) internal information, which measures the counter-party privacy-loss inherent in a communication protocol, (2) output information, which measures the reduction in input-privacy that is inherent when the output of the computation is published, and (3) external information, which measures the privacy-loss inherent when the communication ...
[Full version]

Optimal transportation (OT) has gained considerable popularity as a tool for relating signals defined over geometric domains. Despite recent progress developing generic machinery for understanding and optimizing OT problems, considerable effort is still required to transition OT from a theoretical challenge to a practical tool in the computer graphics, geometry processing, and machine learning toolboxes. To this end, I will describe several efforts to develop efficient, resilient OT-based algorithms tailored to these application areas, including ...
[Full version]

In recent years much work has been devoted to the development of image processing algorithms using local patches. The main idea in this line of work is to impose a statistical prior on the patches of the desired image. An algorithm following this path extracts all possible patches with overlaps from the image, and operates on each separately. The more advanced algorithms exploit also interrelations between different patches in the reconstruction process. In this work ...
[Full version]

In this work we present a new approach of incorporating kernels into dictionary learning. The kernel K-SVD algorithm (KKSVD), which has been introduced recently, shows an improvement in classification performance, with relation to its linear counterpart K-SVD. However, this algorithm requires the storage and handling of a very large kernel matrix, which leads to high computational cost, while also limiting its use to setups with small number of training examples. We address these problems by ...
[Full version]

To recover the three dimensional (3D) volumetric distribution of matter in an object, images of the object are captured from multiple directions and locations. Using these images, tomographic computations seek the distribution. In highly scattering media and constrained irradiance, tomography must explicitly account for off-axis scattering. Furthermore, the tomographic model and recovery must function when imaging is done in-situ, as occurs in medical imaging and ground-based atmospheric sensing. We formulate tomography that handles arbitrary orders ...
[Full version]

The C preprocessor (cpp) is probably the most familiar macro preprocessor. However, it lacks the expressibility of general purpose programming languages. We propose a language approach to preprocessing, based on writing code in a high-level programming language. Our intent is to make it easy to identify structures in source files, and therefore be able to use reflection to some extent. We also suggest a flexible syntax for macro calls, for adding constructs that are more ...
[Full version]

Fluent APIs increasingly gain popularity and research interest. It allows an API designer to enforce correct usage of the API by type-safety. Although many people design new fluent APIs manually, the question of automating the process remained unanswered yet. In this work, we explain why the problem's core lies with the expressive power of Java generics. Our main result is that automatic generation is possible whenever the specification is an instance of the set of ...
[Full version]

Time: 10:00-10:45 Speaker: Prof Myung Soo Kim (Computer Science and Engineering Department, Seoul National University, Korea) Title: Efficient Voronoi Diagram Construction for Planar Freeform Spiral Curves Abstract: We present a real-time algorithm for computing the Voronoi diagram of planar freeform piecewise-spiral curves. The efficiency and robustness of our algorithm is based on a simple topological structure of Voronoi cells for spirals. Using a Möbius transformation, we provide an efficient search for maximal disks. The correct ...
[Full version]

We propose a novel probabilistic framework to model connectivity patterns in the brain as a latent network graph. In particular, we model the interaction between latent anatomical and functional connectivity and present an intuitive extension to population studies. The method simultaneously infers the templates of latent connectivity for each population and the differences in connectivity between the groups. We also develop an approach to identify foci of a neurological disorder based on anatomical and functional ...
[Full version]

n the past several years, the field of computer vision has chalked up significant achievements, fueled by new algorithms (such as deep neural networks), new hardware (such as consumer 3D cameras) and new available processing (such as GPU’s). When we consider the problems that tomorrow’s household robots and autonomous vehicles will have to solve, however, there is evidently still a ways to go. In this talk, I will discuss current work within Intel’s Perceptual Computing ...
[Full version]

Stochastic analysis of real-world signals consists of 3 main parts: mathematical representation; probabilistic modeling; statistical inference. For it to be effective, we need mathematically-principled and practical computational tools that take into consideration not only each of these components by itself but also their interplay. This is especially true for a large class of computer-vision and machine-learning problems that involve certain mathematical structures; the latter may be a property of the data or encoded in the ...
[Full version]

Cloud Computing is considered to be one of the most important paradigm shifts of this century. In this presentation we will explain what cloud computing is all about, review the key drivers and inhibitors of cloud and what is fueling its exponential growth. We will examine its effects across the ecosystem and megatrends as Mobile and Internet Of Things, including the threats and the opportunities. We will also try to determine where we are in ...
[Full version]

Training accurate visual classifiers from large data sets critically depend on learning the right representation for the problem. In this talk, I will discuss a representation learning framework based on an iterative interaction of two components: a feature generator suggesting candidate features, and a feature selector choosing among them. In the feature selector role, I will present a feature selection algorithm for Support Vector Machines (SVMs) enabling selection among hundreds of thousands of features, while ...
[Full version]

CS students in the course "Project in Advanced Programming" invite you to an android applications fair on Wednesday, January 20, 2016, between 14:15-16:30, in room 337 (3rd floor) of CS Taub Building. The fair will exhibit new android applications designed to facilitate life in all aspects and Qualcomm will hold a project competition between the presenting projects: ClimbIt - An application that will send wall-climbers straight to the top. CScourses - An application that helps ...
[Full version]

A probabilistically checkable proof (PCP) enables checking, for example, the satisfiability of a 3-SAT boolean formula phi, while only examining a constant number of randomly chosen symbols of the proof. Suppose an assignment to phi has length n. A long line of research led to the construction of such PCPs of length quasi-linear in n. An additional property of a PCP that can be useful is that of zero-knowledge (ZK): A PCP is zero-knowledge with ...
[Full version]

Graph optimization problems are the most studied problems in theoretical computer science. These problems are not only mathematically intriguing. They have a crucial impact on our increasingly-computerized environment. Among the tractable optimization problems, the most fundamental are shortest paths, maximum flow, minimum cut, maximum matching, and minimum spanning tree. On planar graphs, these problems are all inter-related through fascinating connections. By exploiting these connections, and the remarkable wealth of structural properties exhibited by planar graphs, ...
[Full version]

Correctly matching feature points across images is an important preprocessing step for many computer vision applications (specifically such that require geometric reasoning). Once an initial set of putative matches is obtained, the common methods to detect the correct matches, e.g., RANSAC, use a parametric model that relates the correct matches between an image pair, e.g., the fundamental matrix. To obtain an accurate (uncontaminated) such model, a large number of search iterations is usually required, which ...
[Full version]

CS students in the course "Project in Advanced Programming" invite you to the Winter 2016 Arduino project fair on Sunday, January 17 2016, between 13 00-16.30, in room 337 (3rd floor) of CS Taub Building. The fair will exhibit new Arduino pojects designed to facilitate life in all aspects as follows: Air Hockey - Automatic Air Hockey Table challenge. ArCoffee Table - A table that improves your coffee break in many ways. arDRAWino - How ...
[Full version]

In this research we study the long-standing open question regarding the computational complexity of one of the core problems in supply chains management, the periodic joint replenishment problem. This problem has received a lot of attention over the years and many heuristic and approximation algorithms were suggested. However, in spite of the vast effort, the complexity of the problem remained unresolved. In this paper, we provide a proof that the problem is indeed strongly NP-hard. ...
[Full version]

Traditionally, network measurement takes a small-data approach. Data is expensive, must be gathered unobtrusively, validated carefully, and used to address sharply-defined problems, if we are to obtain reliable answers. The undeniable existence of BigData in and around telephone and data networks (which have merged years ago ) and the presence of tools for learning from unstructured masses of data are changing this. A recent EU tender has called for a “crowdsourcing” approach to characterizing Internet ...
[Full version]

A deterministic preprocessing algorithm, especially designed to deal with repeated structures and wide baseline image pairs, is presented. It generates putative matches and their probabilities. They are then given as input to state-of-the-art epipolar geometry estimation algorithms, improving their results considerably, succeeding on hard cases on which they failed before. The algorithm consists of three steps, whose scope changes from local to global. In the local step, it extracts from a pair of images local ...
[Full version]

Making a Pixar movie involves a team of hundreds of people from widely different disciplines using a lot of sophisticated technology. The scale and complexity of the process make it challenging to create a cohesive story that audiences find moving. This talk looks at how Pixar addresses that challenge. The principals involved are pertinent to any creative endeavor that involves large teams and sophisticated technology, but the visual and emotional nature of movies makes those ...
[Full version]

Pixar’s animated films are created using computer graphics, so the characters are constructed and animated in a virtual 3D world. Manipulating that world involves using physics and math for everything from sculpting the shapes of objects to animating the characters to lighting the scenes to texturing the surfaces to simulating the motion of clothes and hair. And although cartoon physics is based on classical physics, it has to adapt to the wacky things animators do. ...
[Full version]

Traditional combinatorial optimization problems require finding solutions for a single problem instance. However, many real-life applications involve systems that change dynamically over time. Thus, throughout the continuous operation of such a system, it is required to compute solutions for new problem instances, derived from previous instances. Moreover, since the transition from one solution to another incurs some cost, a natural goal is to have the solution for the new instance close to the original one ...
[Full version]

Compositional verification techniques aim to decompose the verification of a large system into the more manageable verification of its components. In recent years, compositional techniques have gained significant successes following a breakthrough in the ability to automate assume-guarantee reasoning. However, automation is still restricted to simple acyclic assume-guarantee rules. In this work, we focus on automating circular assume-guarantee reasoning in which the verification of individual components mutually depends on each other. We use a sound ...
[Full version]

The folklore on quantum entanglement is that it is a fragile phenomenon and physically very hard to maintain. In quantum complexity theory this raises the possibility that there is no analog of the classical hardness-of-approximation in the quantum setting, namely a quantum analog of the PCP theorem. The problem of whether entanglement is robust-enough to even allow a quantum PCP was formalized by Freedman and Hastings ['13] in a conjecture called NLTS (No Low-Energy Trivial ...
[Full version]

The next generation of Intel’s multicore and many-core product lines will use AVX-512, the biggest extension to Intel Instruction Set Architecture (ISA). This extension will provide new HW capabilities, yet poses a critical challenge for SW – how can modern compilers and programmers make efficient use of these new capabilities? This talk will discuss recent research and development advancements in coping with this challenge, in terms of innovative compilation technology, programming models, and the interaction ...
[Full version]

This lecture will address the fundamental question: What are deep neural networks doing to metrics in the data? We know that two important properties of a classification machinery are: (i) the system preserves the important information of the input data; (ii) the training examples convey information for unseen data; and (iii) the system is able to treat differently points from different classes. We show that these fundamental properties are inherited by the architecture of deep ...
[Full version]

While the solution of linear inverse problems (BIPs) under both classical signal models and modern sparsity models has been studied extensively and is well understood, relatively little is known about the solution of bilinear inverse problems. In signal processing, these problems arise in so-called blind signal recovery applications. One notable example is blind deconvolution, with applications in blind image deblurring, blind channel equalization, speech dereverberation, and seismic data analysis. Another important example, is blind gain ...
[Full version]