The Taub Faculty of Computer Science Events and Talks

In the metric compression problem, we are given n points in a metric space, and the goal is to construct a compact representation (sketch)of the points, such that the distance between every pair can be approximate lyre covered from the sketch, up to a small distortion of (1 +/- epsilon). Such sketches are widely used for fast nearest neighbor search in high-dimensional Euclidean spaces. We give a new algorithm for sketching Euclidean metric spaces. It ...
[Full version]

We present the first general and practical solution of the fluent API problem – an algorithm, that given a deterministic language (equivalently, LR(k), k≥0 language) encodes it in an unbounded parametric polymorphism type system employing only a polynomial number of types. The theoretical result is accompanied by an actual tool Fling – a fluent API compiler-compiler in the venue of YACC, tailored for embedding DSLs in Java.
[Full version]

Leveraging Programmable Switches for In-network Computing The network line rate is constantly on the rise to support the exploding amounts of data. This means that we have less time to process individual packets, despite a growing demand for better network telemetry. Moreover, CPU speeds are not rising at the same rate as we near the end of Moore’s law, making it harder to rely on software computations. Programmable hardware switches are an emerging technology that ...
[Full version]

The security and architecture communities will remember 2018 as the year of side channels. Starting from Spectre and Meltdown, time and again we have seen how basic performance-improving features can be exploited to violate fundamental hardware security guarantees. In this talk, I will survey two recent results on hardware side channel attacks. The first, Fallout, continues a series of Meltdown-type vulnerabilities, showing that speculative execution attacks are possible even on post-Meltdown hardware platforms. The second, ...
[Full version]

In recent years, new forms of communication between people and devices have revolutionized our daily lives. The Internet has become the leading platform for human interaction (e.g., social networks), commerce, information, and also control of physical devices (e.g., Internet of Things). This new connectivity creates new security and privacy risks for individual users and organizations. It also increases the complexity and diversity of the different security and cryptographic solutions we need to protect against increasingly ...
[Full version]

In 2010, Patrascu proposed the Multiphase problem, as a candidate for proving polynomial lower bounds on the operational time of dynamic data structures. Patrascu conjectured that any data structure for the Multiphase problem must make n cell-probes in either the update or query phase, and showed that this would imply similar unconditional lower bounds on many important dynamic data structure problems. Alas, there has been almost no progress on this conjecture in the past decade ...
[Full version]

Learning systems have become ever more pervasive in our lives. By controlling the flow of information, they play an increasingly significant role in shaping decisions made by us, about us, and for us. But systems are still developed under paradigms in which human actors are abstracted away, limiting their effectiveness. In this talk I will argue for the need to infuse into machine learning the tangible presence of humans, and present two projects addressing this ...
[Full version]

Approximating ranks, quantiles, and distributions over streaming data is a central task in data analysis and monitoring. Given a stream of n items from a data universe U (equipped with a total order), the task is to compute a sketch (data structure) of size poly(log(n), 1/ε). Given the sketch and a query item y ∈ U, one should be able to approximate its rank in the stream, i.e., the number of stream elements smaller than ...
[Full version]

Existing proofs that deduce P=BPP from circuit lower bounds convert randomized algorithms to deterministic ones with a large polynomial slowdown in running time. In this talk, I will show that if we assume exponential lower bounds against nondeterministic circuits, we can convert any randomized algorithm running in time T to a deterministic one running in time T^{2+����} for an arbitrarily small constant ����. Under complexity-theoretic assumptions, such a slowdown is nearly optimal. Our result follows ...
[Full version]

In recent years, new forms of communication between people and devices have revolutionized our daily lives. The Internet has become the leading platform for human interaction (e.g., social networks), commerce, information, and also control of physical devices (e.g., Internet of Things). This new connectivity creates new security and privacy risks for individual users and organizations. It also increases the complexity and diversity of the different security and cryptographic solutions we need to protect against increasingly ...
[Full version]

There is a recent surge of interest in performing computations on encrypted data. Techniques such as secure multi-party computations and fully-homomorphic encryption enable rich privacy-preserving applications. On the other hand, building such applications is hard due to the cryptographic expertise required to build them correctly, securely, and efficiently. Excitingly, a compiler that raises the level of programming abstraction while performing a slew of domain-specific optimizations can make a huge difference in building these applications. I ...
[Full version]

Although distributed computing can significantly reduce the training time of deep neural networks, scaling the training process while maintaining high efficiency and final accuracy is challenging. Cloud computing, which is becoming increasingly popular as a platform for distributed training of deep neural networks, is prone to straggler workers who significantly increase training time in synchronous environments. Therefore, asynchronous distributed training is preferable in terms of training time when using cloud computing. Distributed asynchronous training enjoys ...
[Full version]

Efficient learning requires prior knowledge (inductivebias). The algorithm designer can manually insert a prior based on hisintuition, but ideally, we would like to automatically infer the mostbeneficial prior. In meta-learning an agent extracts a ‘learned prior’ fromseveral observed learning tasks, which in turn, can be used to facilitate thelearning of new related tasks. The prior should capture the common structureacross learned tasks while allowing sufficient flexibility to adapt tonovel aspects of new tasks. We present ...
[Full version]

Establishing a low network latency connection to stock exchanges has been the desire of trader for years. A well-known example is the extremely expensive construction of an ultra-low latency fiber optic cable between New York and Chicago just for reducing three millisecond in the round trip time. Yet, the exact possible usage and potential profitability of such high end network connection remains mostly unclear. In this Thesis, we address this point by studying the impact ...
[Full version]

Data is one of the most important assets of the information age, and its societal impact is undisputed. Yet, rigorous methods of assessing the quality of data are lacking. In this talk, we propose a formal definition for the quality of a given static dataset. We assess a dataset's quality by a quantity we call the expected diameter, which measures the expected disagreement between two randomly chosen hypotheses that explain it, and has recently found ...
[Full version]

I will speak about an unexpected correspondence between a beautiful theory, due to Lovasz, about homomorphisms and graph limits and a popular heuristic for the graph isomorphism problem known as the Weisfeiler-Leman algorithm. I will also relate this to graph kernels in machine learning. Indeed, the context of this work is to design and understand similarity measures between graphs and discrete structures. (Joint work with Jan Böker, Holger Dell, and Gaurav Rattan.)
[Full version]

Effective data analysis often depends on data that is known to different sources, including private data whose owners cannot disclose. The task at hand is to perform effective analysis of the data while preserving its privacy. This talk will describe efficient cryptographic protocols, some of them based on variants of private set intersection (PSI), that can be applied to perform private analysis of data. Short Bio: ============ Benny Pinkas is a member of the Bar ...
[Full version]

In the 'frequent items' problem one sees a sequence of items in a stream (e.g. a stream of words coming into a search query engine like Google) and wants to report a small list of items containing all frequent items. In the 'change detection' problem one sees two streams, say one from yesterday and one from today, and wants to report a small list of items containing all those whose frequencies changed significantly. For both ...
[Full version]

The PPSZ algorithm, due to Paturi, Pudlak, Saks and Zane, is currently the fastest known algorithm for the k-SAT problem, for every k>3. For 3-SAT, a tiny improvement over PPSZ was obtained by Hertli. We introduce a biased version of the PPSZ algorithm using which we obtain an improvement over PPSZ for every k≥3. For k=3 we also improve on Herli’s result and get a much more noticeable improvement over PPSZ, though still relatively small. ...
[Full version]

In the Bounded Degree Matroid Basis Problem, we are given a matroid and a hypergraph on the same ground set, together with costs for the elements of that set as well as lower and upper bounds f(e) and g(e) for each hyperedge e. The objective is to find a minimum-cost basis B such that f(e) <= |B \cap e| <= g(e) for each hyperedge e. Kiraly, Lau and Singh (Combinatorica, 2012) provided an algorithm that finds a basis of cost at most the optimum value which violates the lower and upper bounds by at most 2\Delta-1, where \Delta is the maximum degree of the hypergraph. We consider an extension of the matroid basis problem to generalized polymatroids, and additionally allow element multiplicities. Building on the approach of Kiraly, Lau and Singh, we provide an algorithm for finding a solution of cost at most the optimum value having the same additive approximation guarantee. As an application, we develop a 1.5-approximation for the metric many-visits TSP, where the goal is to find a minimum-cost tour that visits each city v a positive number r_v of times. Our approach combines our algorithm for the Lower Bounded Degree Generalized Polymatroid Basis Problem with Multiplicities with the principle of Christofides' algorithm from 1976 for the (single-visit) metric TSP, whose approximation guarantee it matches. We also present a new algorithm for the general many-visits TSP without any metric assumptions on the edge cost. For this problem, Cosmadakis and Papadimitriou (SICOMP, 1984) provided an algorithm that finds an optimal tour in time and space that depends superexponentially on the number of cities. We give an improved algorithm which simultaneously improves the time complexity to single-exponential, and the space complexity to polynomial. Assuming the Exponential Time Hypothesis, the run time of our algorithm is asymptotically optimal. Our algorithm is arguably simpler than the one by Cosmadakis and Papadimitriou. (Based on joint work with Kristof Berczi, Andre Berger, Tamas Kiraly, Laszlo Kozma, Gyula Pap and Roland Vincze) Short Bio: ========== Matthias Mnich is an assistant professor of quantitative economics at Maastricht University, The Netherlands, currently on leave for research in theoretical computer science at the Rheinische Friedrich-Wilhelms Universitaet Bonn, Germany. He obtained his PhD in 2010 at Eindhoven University of Technology, for which he received the Philips Prize 2010. Since then he held positions at UC Berkeley, USA and the Max Planck Institute for Computer Science, and a visiting professorship at TU Darmstadt, Germany. His research interests are in combinatorial algorithms, social choice theory, algorithms for big data, and algorithms in artificial intelligence. ================================================== Refreshments will be served from 14:15 Lecture starts at 14:30
[Full version]

We present a method for determining which facial parts (mouth, nose, etc.) best characterize an individual, given a set of that individual's portraits. We introduce a novel distinctiveness analysis of a set of portraits, which leverages the deep features extracted by a pre‐trained face recognition CNN and a hair segmentation FCN, in the context of a weakly supervised metric learning scheme. Our analysis enables the generation of a polarized class activation map (PCAM) for an ...
[Full version]

Consider an instance of Euclidean k-means or k-medians clustering. We show that the cost of the optimal solution is preserved up to a factor of (1+ε) under a projection onto a random - dimensional subspace. Further, the cost of every clustering is preserved within (1+ε). Our bound on the dimension is nearly optimal. Additionally, our result applies to Euclidean k-clustering with the distances raised to the p-th power for any constant p. Dr. Makarychev is ...
[Full version]

Interactive oracle proofs are a hybrid between interactive proofsand probabilistically-checkable proofs, where the prover is allowed to interactwith a verifier by sending relatively long messages to the verifier, who inturn is only allowed to query a few of the bits that were sent. In this work we construct, for any NP relation for whichmembership can be decided in polynomial-time and bounded polynomial space (e.g.,SAT, Hamiltonicity, Clique, Vertex-Cover, etc.), interactive oracle proofs inwhich the proof length ...
[Full version]

The Network Time Protocol (NTP) synchronizes time across computer systems over the Internet. Unfortunately, NTP is highly vulnerable to “time shifting attacks”, in which the attacker’s goal is to shift forward/backward the local time at an NTP client. NTP’s security vulnerabilities have severe implications for time-sensitive applications and for security mechanisms, including TLS certificates, DNS and DNSSEC, RPKI, Kerberos, BitCoin, and beyond. While technically NTP supports cryptographic authentication, it is very rarely used in practice ...
[Full version]

Photos and videos are now a main mode of communication, used to tell stories, share experiences and convey ideas. However, common media editing tools are often either too complex to master, or oversimplified and limited. In this talk I will present my strategy towards the creation of media editing techniques that are easy to learn, yet expressive enough to reflect unique creative objectives. We will mostly discuss one specific domain --- human heads --- which ...
[Full version]

SINR (aka physical model) is a model of wireless communication that has been studied from algorithmic perspective in the last decade, and is considered as a more realistic alternative to the traditional graph-based models. I will introduce the model and discuss some basic scheduling problems. At a high level, these can be thought of as coloring or independent set problems in some special fractional graphs.
[Full version]

SpaceIL made history when Beresheet, the first privately funded spacecraft, have reached the Moon with a minute budget of 100M$. After eight years of planning, Beresheet's journey started on February 22nd, 2019, when SpaceIL's lunar lander has been launched to space on top of a Falcon 9 launcher. During its 49 days space journey, Beresheet performed a series of successful autonomous maneuvers that brought it from Earth orbit to lunar orbit, and on April 11th, ...
[Full version]

This paper defines the notion of class discrepancy for families of functions. It shows that low discrepancy classes admit small offline and streaming coresets. We provide general techniques for bounding the class discrepancy of machine learning problems. As corollaries of the general technique we bound the discrepancy (and therefore coreset complexity) of logistic regression, sigmoid activation loss, matrix covariance, kernel density and any analytic function of the dot product or the squared distance. Our results ...
[Full version]

Traditional query optimizer is crucially dependent on cardinality estimation, which enables choosing among different plan alternatives by using the cardinality estimation of intermediate results within query execution plans. Therefore, the query optimizer must use reasonably good estimates. However, estimates produced by all widely-used database cardinality estimation models are routinely significantly wrong, resulting in not choosing the best plans, leading to slow executions. In this work we define a new problem, that of estimating the containment ...
[Full version]

The overlapping correlation clustering problem has multiple real-world applications in diverse areas such as: social networks (community detection), biology (protein analysis) and information retrieval (categorising documents). Correlation clustering is a problem of assigning data points into distinct groups (clusters) based on similarity. When assigning two similar data points to the same cluster or assigning two different data points to different clusters, we score this assignment as successful, otherwise, we score the assignment as unsuccessful. When ...
[Full version]

Lifting theorems are theorems that relate the query complexity of a function $f : \{0, 1\}^{n} \rightarrow \{0, 1\}$  to the communication complexity of the composed function fgn, for some “gadget” $g  : \{0, 1\}^{b} \times \{0, 1\}^{b} \rightarrow \{0, 1\}$. Such theorems allow transferring lower bounds from query complexity to the communication complexity, and have seen numerous applications in the recent years. In addition, such theorems can be viewed as a strong generalization of ...
[Full version]

When this logician moved to theoretical computer science, this did not seem such a big deal to him. The transition was seamless. The transition to software engineering was radical. Microsoft research invited me to come, start a group and use my ideas to build tools. That was enticing and frightening. The chances of success seemed remote. But it worked, thanks to the engineering talent that my group was lucky to attract. Our Spec Explorer became ...
[Full version]

Reparameterization of variational auto-encoders with continuous random variables is an effective method for reducing the variance of their gradient estimates. Our work optimizes the discrete VAE objective directly, using its Gumbel-Max reparameterization. This optimization technique propagates gradients through the reparameterized argmax, and replaces MCMC sampling by the difference of gradients of two argmax predictions. This realization provides the means to learn latent representations in cases when evaluating the argmax operation is tractable while evaluating the ...
[Full version]

Learning to generate music is an ongoing AI challenge. A more difficult challenge is the generation of musical pieces that match human-specific preferences. In this work we focus on personalized, symbol-based, monophonic generation of harmony-constrained jazz improvisations. To tackle this objective, we introduce a pipeline consisting of the following steps: supervised learning using a corpus of solos, high-resolution user preference metric learning, and optimized generation using planning (beam search). Our corpus consists of hundreds of ...
[Full version]

Suppose we receive a stream of data, sampling each arriving data element with some predetermined probability. It is well known that if the elements in the stream are chosen in advance, then a constant number of random samples suffice to make the sampled subset “represent” the whole stream with good probability. However, what if the elements are chosen adaptively, by an adversary that knows exactly, at any point along the stream, which elements were sampled ...
[Full version]

Artificial neural networks are nowadays routinely trained to solve challenging learning tasks, but our theoretical understanding of this phenomenon remains quite limited. One increasingly popular approach, which is aligned with practice, is to study how making the network sufficiently large (a.k.a. ''over-parameterized'') makes the associated training problem easier. In this talk, I'll describe some of the possibilities and challenges in understanding neural networks using this approach. Based on joint works with Itay Safran and Gilad ...
[Full version]

In many common datasets, neural networks can achieve zero training loss yet generalize well to unseen data. Recent works suggest that this because standard training algorithms (e.g., GD or SGD) have an implicit regularization which is biased towards specific solutions, which tend to have good generalization. I will review such "algorithmic biases", how they affect the functional capabilities of the neural networks, and how this relates to generalization in simple models.
[Full version]

How does the brain beget the mind? How do molecules, cells and synapses effect reasoning, intelligence, language, science? Despite dazzling progress in experimental neuroscience we do not seem to be making progress in the overarching question -- the gap is huge and a completely new approach seems to be required. As Richard Axel recently put it: "We don't have a logic for the transformation of neural activity into thought." What kind of formal system would ...
[Full version]

This work is dedicated to eliminating the overhead required for guaranteeing the storage order in the modern IO stack. The existing block device adopts a prohibitively expensive approach in ensuring the storage order among write requests: interleaving the write requests with Transfer-and-Flush. Exploiting the cache barrier command for Flash storage, we overhaul the IO scheduler, the dispatch module, and the filesystem so that these layers are orchestrated to preserve the ordering condition imposed by the ...
[Full version]

The talk is about anti-concentration of the inner product of two independent random vectors in Euclidean space. We shall discuss a proof, and some applications
[Full version]

Technion CS celebrates jubilee and invites you to CS City - a festival of science and technology for the whole family which will be held on Tuesday, October 29, 2019 starting at 4:30 pm at the CS Taub Building, Technion. Participation is free, but pre-registration is required. You are all invited!
[Full version]

Election databases are the main elements of a recently introduced framework that aims to create bridges between the computational social choice and the data management communities. An election database consists of incomplete information about the preferences of voters, in the form of partial orders, alongside with standard database relations that provide contextual information. Earlier work in computational social choice focused on the computation of possible winners and necessary winners that are determined by the available ...
[Full version]

Managing data inconsistency has been one of the major challenges in the research and practice of database management. Database inconsistency arises for different reasons and in different applications. Nowadays, many applications obtain information from imprecise sources (e.g., social encyclopedias, social networks, and sensors attached to appliances) via imprecise procedures (e.g., natural-language processing, signal processing, and image analysis). Inconsistency may also arise when integrating conflicting data from different (possibly consistent) sources. During the past two decades, ...
[Full version]

A tree code is a combinatorial object introduced by Schulman in the early 90s as a key ingredient in interactive coding. Schulman proved the existence of tree codes with constant alphabet size, but the problem of explicitly constructing a tree code remained elusive. We present an explicit binary tree code with constant distance and alphabet size polylog(n), where n is the depth of the tree. This improves over the (Evans-Klugerman-Schulman 1994) construction that requires an ...
[Full version]

Inverse problems in imaging are extensively studied, with a variety of strategies, tools, and theory that have been accumulated over the years. Recently, this field has been immensely influenced by the emergence of deep-learning techniques. One such contribution, which is the focus of this paper, is the Deep Image Prior (DIP) work by Ulyanov, Vedaldi, and Lempitsky (2018). DIP offers a new approach towards the regularization of inverse problems, obtained by forcing the recovered image ...
[Full version]

Large training datasets are required to achieve competitive performance in most natural language tasks. The acquisition process for these datasets is labor intensive, expensive, and time consuming. This process is also prone to human errors. In this work, we show that cross-cultural differences can be harnessed for natural language text classification. We present a transfer-learning framework that leverages widely-available unaligned bilingual corpora for classification tasks, using no task-specific data. Our empirical evaluation on two tasks ...
[Full version]

Product descriptions play an important role in the e-commerce ecosystem, conveying information to buyers about merchandise they may purchase. Yet, on leading e-commerce websites, with high volumes of new items offered for sale every day, product descriptions are often lacking or missing altogether, and when they do appear, they lack personalization to the user. We suggest to mitigate these issues by generating short crowd-based product descriptions from user reviews and then explore how to personalize ...
[Full version]

Maintaining an accurate trained model on an infinite data stream is challenging due to concept drifts that render a learned model inaccurate. Updating the model periodically can be expensive, and so traditional approaches for computationally limited devices involve a variation of online or incremental learning, which tend to be less robust. The advent of heterogeneous architectures and Internet-connected devices gives rise to a new opportunity. A weak processor can call upon a stronger processor or ...
[Full version]

On Wednesday, September 11, TCE center will host this year the 9th annual Henry Taub International Conference on Autonomous Systems and will focus on the advent of autonomous systems including self-driving cars, automatic delivery drones, and service robots considered by many a major revolution of modern times that holds a profound impact on the economy and society. The event will be a unique opportunity to hear and meet international experts from the academia and the ...
[Full version]

Non-volatile memory is expected to co-exist or replace DRAM in upcoming architectures. Durable concurrent data structures for non-volatile memories are essential building blocks for constructing adequate software for use with these architectures. In this paper, we propose a new approach for durable concurrent sets and use this approach to build the most efficient durable hash tables available today. Evaluation shows a performance improvement factor of up to \maxImprovement{} over existing technology.
[Full version]

We construct a new interactive polynomial commitment scheme using groups of unknown order. The interactive evaluation argument is public-coin and has logarithmic communication in the degree of the polynomial. The scheme does not require a trusted setup if instantiated with class groups. We apply this as a cryptographic compiler to polynomial IOPs in order to obtain preprocessing public-coin interactive arguments of knowledge with succinct online communication (without trusted setup). Instantiated with the IOP from Sonic ...
[Full version]

In the field of population genetics, it is known that collected observations are not independent, as they all share a common ancestor. Current state-of-the-art association tests use Monte Carlo simulations to account for these dependency structures. However, these simulations are not scalable and are cumbersome to use for the evaluation of only a few parameters. In our work, we introduce a generalized, simulation-free analytical test that accounts for hierarchical sample dependency structures. We formulate our ...
[Full version]

The purpose of this research is to explore an innovative approach to the declarative and imperative paradigms of programming languages. To demonstrate this approach, we have developed a prototype domain-specific language inspired by the scripts of theatrical plays, hence dubbed Tʜᴀᴘʟ. Tʜᴀᴘʟ's intended use is in the context of animation generation in “slide-show” presentations. There are two different ways to integrate animations in a presentation. The first is to embed an animation within a single ...
[Full version]

Massively parallel, throughput-oriented processors are becoming increasingly common. Maximizing the benefit of these processors requires algorithms to be implemented differently than the sequential algorithms that most software developers are familiar with. This change is often very time-consuming and is not guaranteed to give an increase in performance matching the amount of effort. This work proposes an approach to predict the performance gain from porting an algorithm from the CPU to the GPU, based only on ...
[Full version]

Real-world networks are dynamic in nature -- nodes may join or leave the network at any time, and communication links may appear or disappear constantly. In this dynamic setting, we study the problem of triangle detection, where the nodes of a network need to determine, distributively, whether or not the network contains a triangle. We examine several variants of this problem, and present algorithms and impossibility results. Then, we extend our results to larger cliques.
[Full version]

Concurrent data structures provide fundamental building blocks for concurrent programming. Standard concurrent data structures may be extended by allowing a sequence of operations to be submitted as a batch for later execution. A sequence of such operations can then be executed more efficiently than the standard execution of one operation at a time. We propose a novel algorithmic extension to the prevalent FIFO queue data structure that exploits such batching scenarios. An implementation in C++ ...
[Full version]

Numerical geometry comprises of principled computational methods that utilize theoretical insights from geometry along with the engineering concepts from numerical methods, for tackling various problems in geometric data analysis. In contrast, computational methods from recent advances in deep learning exhibit a black box nature where essential and meaningful features are learned from examples of training data leading to state-of-the-art results. This thesis explores a synergy between these two disparate computational philosophies. In particular, we integrate ...
[Full version]

Jigsaw puzzle solving is an intriguing problem, having a wide variety of applications in archaeology, biology, document restoration and others. This paper focuses on a specific variant of the problem - solving puzzles with eroded boundaries. Such erosion makes the problem extremely difficult, since most existing solvers utilize solely the information at the boundaries. Nevertheless, this variant is important since erosion and missing data often occur at the boundaries. This paper introduces a novel approach ...
[Full version]

Suboptimal resource utilization among public and private cloud providers prevents them from maximizing their economic potential. Long-term allocated resources are often idle when they might have been subleased for a short period. Alternatively, arbitrary resource overcommitment may lead to unpredictable client performance. We propose a mechanism for fixed availability (traditional) resource allocation alongside stochastic resource allocation in the form of shares. We show its benefit for private and public cloud providers and for a wide ...
[Full version]

The information in the weights of deep networks plays a key role in understanding their behavior. When used as a regularizer during training (either explicitly or implicitly) it induces generalization through the PAC-Bayes bound. Rather than being computed and minimized explicitly, it can be directly controlled by injecting noise in the weights, a process known as Information Dropout. It can also be shown that stochastic gradient descent (SGD), when taken to the continuous limit and ...
[Full version]

Recent years have seen great progress in automated synthesis techniques that can automatically generate code based on some intent expressed by the user, but communicating this intent remains a major challenge. When the expressed intent is coarse-grained (for example, restriction on the expected type of an expression), the synthesizer often produces a long list of results for the user to choose from, shifting the heavy-lifting to the user. An alternative approach is programming by example ...
[Full version]

The omnipresence of increasingly large deep networks derives primarily from their empirical successes. Indeed, empirical evidence places a strong emphasis on operating at scale, more parameters and layers, in order to help both optimization and generalization. This counter-intuitive trend has eluded theoretical analysis. In this work, we present a PAC-Bayesian generalization bound for the Gaussian dropout that only requires an on-average loss function bound and on-average norm-gradient bounds by relying on log-Sobolev inequalities for Gaussian ...
[Full version]

Different methods where presented through the years to find good solutions for the Matrix Completion Problem. This problem appears in many tasks in life where a sparse signal, which lies on a grid of two non- Euclidean domains (Graphs or manifolds), should be predicted or completed. A classic example of this problem is the “Netflix Problem” which appears in the field of “Recommendation Systems” or “recommender systems”. In those systems recommendations of specific items are ...
[Full version]

This talk will present our approach for enumerating the results of queries in monadic second-order (MSO) on trees. Our goal is to first preprocess the input tree in linear time, and then enumerate the answers with delay linear in the size of each answer, i.e., constant-delay for queries with free first-order variables. Our approach is also tractable in the query when represented by a tree automaton (not necessarily deterministic). The approach works by computing a ...
[Full version]

Applications' memory consumption is growing rapidly, which increases their dependence on the memory system. Since each virtual-to-physical memory translation in modern systems requires cache lookups or multiple memory accesses, the run time overhead of virtual memory translation is on the rise. Meanwhile, the increasing prevalence of the cloud ecosystem and server consolidation moves more applications to run in virtual machines. This further aggravates the overhead of memory translation; in virtual machines, the processor needs to ...
[Full version]

Whereas deterministic protocols are typically guaranteed to obtain particular goals of interest, probabilistic protocols typically provide only probabilistic guarantees. In this talk we initiate an investigation of the interdependence between actions and subjective beliefs of agents in a probabilistic setting. In particular, we study what probabilistic beliefs an agent should have when performing actions, in a protocol that must satisfy a given probabilistic constraint. It is shown that the agent need not believe that a ...
[Full version]

Rank-metric codes (RMCs) over finite fields were introduced some 40 years ago by Delsarte, Gabidulin and Roth. Interest in RMCs was revived in recent years due to the application of such codes to network coding. In 1991, another simple construction of a class of RMCs was presented by Roth, which is optimal when the field is algebraically closed. These codes are based on diagonal interleaving of maximum-distance separable (MDS) codes. In 2017, a decoding scheme ...
[Full version]

Rigid registration is the problem of finding the optimal rigid motion and correspondence between two shapes, so that they are as similar as possible in terms of an appropriate energy. We will describe several popular algorithms for this problem: PCA alignment and ICP, which are very efficient but are not globally optimal, as well as sampling and branch and bound (BnB) algorithms which exhibit slow convergence but are globally optimal. Next we suggest our quasi ...
[Full version]

In this talk, we will look at decoding Reed-Muller codes beyond their minimum distance when the errors are random (i.e., in the binary symmetric channel). A beautiful result of Saptharishi, Shpilka and Volk showed that for binary Reed-Muller codes of length n and degree n - O(1), one can correct polylog(n) random errors in poly(n) time (which is well beyond the worst-case error tolerance of O(1)). We give two new algorithms for this problem, both ...
[Full version]

Shape correspondence is a fundamental task in shape analysis. Given two shapes, the goal is to compute a semantic correspondence between points on them. Shape correspondence is required when two shapes are analyzed jointly, which is common in many applications such as texture and deformation transfer, statistical shape analysis and shape classification. Automatically computing shape correspondence is a difficult problem, especially when the shapes are significantly different. In this talk I will present a fully ...
[Full version]

Modern cloud architectures rely on each server running a software-defined networking stack to implement policy and virtualization. However, these networking stacks are becoming increasingly complex as features are added and as network speeds increase. Running these stacks on CPU cores takes away processing power from VMs, increasing the cost of running cloud services, and adding latency and variability to network performance, which has led us to invest in technologies to accelerate SDN in NIC hardware ...
[Full version]

It is widely known that in the well-studied \LOCAL model of distributed computing, introduced by Linial [FOCS '87], many of the classic distributed graph problems (including maximal independent set (MIS) and MaxDegree+1-vertex coloring), have very efficient (and simple) randomized but exponentially slower deterministic algorithms. Understanding and potentially narrowing down this exponential gap is considered to be one of the central long-standing open questions in the area of distributed graph algorithms. We present the recent advances ...
[Full version]

A fundamental question in learning to classify 3D shapes is how to treat the data in a way that would allow us to construct efficient and accurate geometric processing and analysis procedures. Here, we restrict ourselves to networks that operate on point clouds. There were several attempts to treat point clouds as non-structured data sets by which a neural network is trained to extract discriminative properties. The idea of using 3D coordinates as class identifiers ...
[Full version]

CS Labs: Systems and Software Development Laboratory (SSDL), Cyber and Computer Security Laboratory (CYBER), The Laboratory for Computer Communication and Networking (LCCN) invite you to visit the Spring Project Fair in IoT, Software, Android Apps, AI, Cyber, Computer Security, and Networks, including demos and presentations by 60 undergraduate teams who will answer your questions on their research. The event will be held on Tuesday, June 25, 2019, at 12, in the CS Taub Lobby. You ...
[Full version]

The 9th CS Research Day for graduate studies will be held on Monday, June 24, 2019 between 15:00-17:00, at the lobby of the CS Taub Building. Research Day events are opportunity for our graduate students to expose their researches using posters and presentations to CS faculty and all degrees students, Technion distinguished representatives and to high-ranking delegates from the hi-tech leading industry companies in Israel and abroad. The participating researches will be on various topics: ...
[Full version]

Datacenter congestion control remains to be a hard problem. The root of the problem is that the ability of end-hosts to react to congestion is fundamentally limited by the timeliness and precision of congestion signals from the network. Unfortunately, despite decades of research, the community is still in quest of such timely and precise congestion signals. In this talk, I will discuss a possible new approach to datacenter congestion control problem: designing network fabrics that ...
[Full version]

In this talk, a new family of codes called clustering-correcting codes is presented. This family of codes is motivated by the special structure of data that is stored in DNA-based storage systems. The data stored in these systems has the form of unordered sequences, also called strands, and every strand is synthesized thousands to millions of times, where some of these copies are read back during sequencing. Due to the unordered structure of the strands, ...
[Full version]

I'll discuss our novel discrete scheme for simulating viscous thin films in 2D at real-time frame rates. Our scheme is based on a new formulation of the gradient flow approach, that leads to a discretization based on local stencils that are easily computable on the GPU. I'll also discuss our recent work toward a real-time 3D scheme for simulation of the effect on meshes, based on simplifications of a previous scheme.
[Full version]

Abstract: One of the main tasks in three-dimensional shape analysis is to retrieve similarities between two or more non-rigid objects in terms of point-to-point correspondences. For the descriptor based approach, it is useful to construct a simplified shape representation called feature descriptor. A successful descriptor class can be motivated by physical phenomena governed by partial differential equations (PDEs). The talk shows how different types of PDEs and discretization aspects may lead to quality improvements for ...
[Full version]

We develop a new approach for distributed distance computation in planar graphs that is based on a variant of the metric compression problem recently introduced by Abboud et al. [SODA'18]. In our variant of the Planar Graph Metric Compression Problem, one is given an $n$-vertex planar graph $G=(V,E)$, a set of $S \subseteq V$ source terminals lying on a single face, and a subset of target terminals $T \subseteq V$. The goal is to compactly ...
[Full version]

Approximating high-dimensional functionals with low-dimensional models is a central issue of machine learning, image processing, physics and mathematics. Deep convolutional networks are able to approximate such functionals over a wide range of applications. This talk shows that these computational architectures take advantage of scale separation, symmetries and sparse representations. We introduce simplified architectures which can be analyzed mathematically. Scale separations is performed with wavelets and scale interactions are captured through phase coherence. We show applications ...
[Full version]

The new and highly effective techniques frequently associated with Artificial Intelligence, including machine learning, pattern recognition, natural language processing, robotics, image analysis, and nonlinear optimization will have profound effects on the ways research is done and the sorts of projects that are feasible. There is a broad spectrum of approaches with differing risks and experience, ranging from • using machine learning as effective tools for extracting patterns from text, sounds, and image data • using ...
[Full version]

In this talk, we consider the problem of communicating over a discrete memoryless two-way channel using non-adaptive schemes, under a zero probability of error criterion. We derive inner and outer bounds on the zero-error capacity region, based on random coding, linear programming, and linear codes.
[Full version]

I will present the 'virtual democracy' framework for the design of ethical AI. In a nutshell, the framework consists of three steps: first, collect preferences from voters on example dilemmas; second, learn models of their preferences, which generalize to any (previously unseen) dilemma; and third, at runtime, predict the voters' preferences on the current dilemma, and aggregate these virtual 'votes' using a voting rule to reach a decision. I will focus on two instantiations of ...
[Full version]

Many natural language understanding (NLU) tasks consist of identifying the relationship between two objects, such as a paragraph and a question (reading comprehension), an image and a question (visual question answering), or a premise and a hypothesis (natural language inference). These tasks supposedly require a deep understanding of the information in the two objects and inference of the relationship between them. However, recent work has demonstrated that many NLU datasets contain one-sided biases—artifacts that allow ...
[Full version]

In recent years, cloud providers have been moving towards offering their clients separate cloud resources for short periods of time instead of offering bundles of resources for longer periods. In parallel, the providers are moving towards using economic mechanisms, such as auctions, to allocate these resources. Vickrey-Clarke-Groves (VCG) auctions are likely to be used for that purpose. These auctions are incentive compatible: by allocating the resource to the highest bidders first, they maximize social welfare---the ...
[Full version]

Recent advances in manufacturing techniques have greatly expanded the types and complexity of shapes that can be produced. At the same time, advances in AI (esp. machine learning) and physics driven algorithms (e.g., topology optimization) have enabled novel methods and workflows for design engineering. A new class of techniques collectively called Generative Engineering are being developed to revamp traditional design and manufacturing systems to enable industries to leverage the full capabilities of new manufacturing technology ...
[Full version]

You are invited to a workshop on Image Processing, Computer Vision, Computer Graphics and Virtual Reality which will be held on Thursday, June 13, 2019, in Room 337, 09:00-12:20, CS Taub Building. More details and full program in the attached poster Please pre-register. You are all invited.
[Full version]

This work provides an additional step in the theoretical understanding of neural networks. We consider neural networks with one hidden layer and show that when learning symmetric functions, one can choose initial conditions so that standard SGD training efficiently produces generalization guarantees. We empirically verify this and show that this does not hold when the initial conditions are chosen at random. The proof of convergence investigates the interaction between the two layers of the network. ...
[Full version]

In this talk, I will describe several research projects developed in the Berkeley Architecture Research group which utilize FPGA-accelerated simulation and open-source hardware, demonstrating the possibilities and accessibility of full-system computer architecture research using these tools and methodologies. FireSim is an easy-to-use, open-source, FPGA-accelerated cycle-accurate hardware simulation platform that runs on Amazon EC2 F1. FireSim automatically transforms and instruments open-hardware designs (e.g. RISC-V Rocket Chip, BOOM, Hwacha, NVDLA, etc.) with the MIDAS framework into fast ...
[Full version]

The last decade saw a tremendous empirical success of machine learning algorithms, which often can not be explained theoretically using a worst-case analysis. Therefore, it is a great challenge to identify reasonable assumptions under which the theory is aligned with the practice. In this talk I will advocate learning in an oracle model, which provides an elegant and extremely useful methodology for this purpose. Namely, instead of explicitly characterizing some desired "niceness" properties of the ...
[Full version]

Reactive synthesis is an automated procedure to obtain a correct-by-construction reactive system from a given declarative, temporal specification. Examples of these systems include the software controllers of robotic systems. Despite recent advancements on the theory and algorithms of reactive synthesis, e.g., efficient synthesis for the GR(1) fragment of linear temporal logic, many challenges remain in bringing reactive synthesis technologies to the hands of software engineers. The SYNTECH project is about bridging this gap. It addresses ...
[Full version]

We present a Monte Carlo rendering framework for the physically-accurate simulation of speckle patterns arising from volumetric scattering of coherent waves. These noise-like patterns are characterized by strong statistical properties, such as the so-called memory effect. These properties are at the core of imaging techniques for applications as diverse as tissue imaging, motion tracking, and non-line-of-sight imaging. Our rendering framework can replicate these properties computationally, in a way that is orders of magnitude more efficient ...
[Full version]

3D objects, modeled using Computer Aided Geometric Design (CAGD) tools, are traditionally represented using a boundary representation (B-rep), and typically use spline functions to parameterize these boundary surfaces. However, recent development in physical analysis, in isogeometric analysis (IGA) in specific, necessitates a volumetric parametrization of the interior of the object. IGA is performed directly by integrating over the spline spaces of the volumetric spline representation of the object. Typically, tensor-product B-spline trivariates are used to ...
[Full version]

Blockchains are distributed secure ledgers to which transactions are issued continuously and each block of transactions is tightly coupled to its predecessors. Permissioned blockchains place special emphasis on transactions throughput. In this paper we present TOY, which leverages the iterative nature of blockchains in order to improve their throughput in optimistic execution scenarios. TOY trades latency for throughput in the sense that in TOY the last f + 1 blocks of each node’s blockchain are ...
[Full version]

Containers (or Operating System based Virtualization) are an old technology; however, the current excitement (and consequent investment) around containers provides interesting avenues for research on updating the way we build and manage container technology. The most active area of research today, thanks to concerns raised by groups supporting other types of virtualization, is in improving the security properties of containers. The first step in improving security is actually being able to measure it in the ...
[Full version]

Arguably the most common network communication problem is multiple-unicasts: Distinct packets at different nodes in a network need to be delivered to a destination specific to each packet, as fast as possible. The famous multiple-unicast conjecture posits that, for this natural problem, there is no performance gap between routing and network coding, at least in terms of throughput. We study the same network coding gap, but in terms of completion time. While throughput corresponds to ...
[Full version]

In the last six years, Bitcoin has obtained a good deal of attention, reaching in May 2019 a market capitalization of more than one hundred and fifty billion (US) dollars and a daily trading volume of hundreds of millions of dollars. Although Bitcoin is considered mostly as a financial asset, the protocol published in 2008 by the anonymous Satoshi Nakamoto goes far beyond: it achieves consensus in an open, decentralized and trust-less network. In this ...
[Full version]

Visual acuity (VA) tests are widely used in ophthalmology to measure vision quality, but even modern tests still rely on an algorithm created in 1976 \cite{Bailey1976}. This algorithm is basically the original VA test algorithm from 1873 \cite{Snellen1873}, with minor modifications. The tests measure VA by requiring the examinee to identify different letters from a table that is positioned in a fixed distance. In this thesis, we present a novel way to assess VA by ...
[Full version]

Computation integrity (CI) protocols allow a strong prover to convince a skeptic verifier it has knowledge of an input satisfying a program. Although theoretical CI constructions are well known for almost 30 years, industrial adaption has only recently started. In this seminar we will go over recent progress in industrial adaption, challenges addressed by our work, and the connection to blockchain based systems.
[Full version]

Deep neural networks (DNNs) have recently shown great success in many machine learning domains and problems. However, applying these models for mission critical tasks still has many safety issues due to prediction uncertainty and prediction errors. This talk covers several related results concerning uncertainty estimation, selective classification (also known as classification with reject option) and active learning for deep neural networks. We discuss first a selective classification method that is based on threshold over an ...
[Full version]

The sequence reconstruction problem corresponds to a model in which a sequence from some code is transmitted over several noisy channels. The channels are almost independent as it is only required that their outputs are different. The main problem under this paradigm is to determine the minimum number of channels required to reconstruct the transmitted sequence. This problem is equivalent to finding the maximum intersection size between two balls of any possible two inputs, where ...
[Full version]

Manufacturing techniques such as hot-wire cutting, wire-EDM, wire-saw cutting, and flank CNC machining all belong to a class of processes called line-cutting where the cutting tool moves tangentially along the reference geometry. From a geometric point of view, line-cutting brings a unique set of challenges in guaranteeing that the process is collision-free. In this work, given a set of cut-paths on a freeform geometry as the input, we propose a conservative algorithm for finding collision-free ...
[Full version]

I will discuss our recent shortest paths algorithms in the distributed congested clique model. Our first contribution is a (2+epsilon)-approximation algorithm for all-pairs shortest paths (APSP) requiring only polylogarithmic number of rounds. This is the first sub-polynomial constant approximation for APSP in this model. We also significantly improve upon the state-of-the-art for approximate multi-source shortest paths and diameter, as well as exact single-source shortest paths (SSSP). Our main techniques are new distance tools that are ...
[Full version]

Graph processing is typically considered to be a memory-bound rather than compute-bound problem. One common line of thought is that more available memory bandwidth corresponds to better graph processing performance. However, in this talk, I will show that this is not necessarily the case. I will demonstrate that the key factor in the utilization of the memory system for graph algorithms is not the raw bandwidth, or even latency of memory requests, but instead is ...
[Full version]

Artificial neural networks have recently revolutionized the field of machine learning, demonstrating striking empirical success on tasks such as image understanding, speech recognition and natural language processing. However, we still do not have sufficient theoretical understanding of how such models can be successfully learned. Two specific questions in this context are: how can neural nets be learned despite the non-convexity of the learning problem, and how can they generalize well despite often having more parameters ...
[Full version]

A prominent property of natural images is that groups of similar patches within them tend to lie on low-dimensional subspaces. This property has been previously used for image denoising, with particularly notable success via weighted nuclear norm minimization (WNNM). In this work, we extend the WNNM method into a general image restoration algorithm, capable of handling arbitrary degradations (e.g. blur, missing pixels, etc.). Our approach is based on a novel regularization term which simultaneously penalizes ...
[Full version]

The aim of photometric stereo is to estimate shape and appearance of a three-dimensional object from multiple input images taken from the same point of view but under different lighting conditions. The most common techniques are conceptually close to the classic photometric stereo problem, meaning that the modelling encompasses a linearisation step and that the shape information is computed in terms of surface normals. Instead of linearising we aim to stick to the original formulation ...
[Full version]

A robust positioning pattern is a large array that allows a mobile device to locate its position by reading a possibly corrupted small window around it. In this talk, we focus on the 1-dimensional case, i.e., robust positioning sequences (RPS). We present constructions of RPSs along with efficient locating algorithms. In particular, we construct a class of q-ary RPSs which are robust to a constant fraction of errors and have asymptotically optimal rate. We also ...
[Full version]

In the classical derandomization problem, we are given a description of a Boolean circuit, and want to deterministically approximate its acceptance probability, up to a small constant error (say, 1/6). This problem is intimately related to circuit lower bounds, and extensive efforts have been devoted in the last four decades to solving it, even for very weak circuit classes. This survey talk is focused on a relaxed version of the problem, which was introduced a ...
[Full version]

Instead of using full cycle-accurate simulations, recent virtual memory studies frequently utilize linear models to predict application runtimes from TLB misses. The benefit of the latter methodology is that it is easier and much faster than the former, but its correctness has never been validated. In fact, previous studies were not even able to measure their prediction errors, because they defined the linear models by collecting at most two empirical data points: the application runtimes ...
[Full version]

Dynamically-typed programming languages such as Python, Ruby and JavaScript are widely used, and much effort is spent on making them efficient. One substantial research effort in this direction is the enabling of parallel code execution. As a first step towards efficiently synchronizing such languages, we look at concurrent data-structures that may modify their memory layout dynamically (e.g. increase their size or change their shape). In this talk we propose a novel Layout Lock that incurs ...
[Full version]

Hyperspectral imaging allows measuring the spectrum of each pixel in the image of a scene. This modality has diverse applications, ranging from agriculture and geoscience, to biomedical imaging and surveillance. However, most existing methods require dedicated cameras, which make use of extra optical elements (prisms, fibers, lenslet arrays, etc.). These solutions are often expensive and cumbersome to deploy in many settings. In this work, we introduce a simple hyperspectral imaging approach, which uses a conventional ...
[Full version]

In this talk, we consider locally repairable codes which are optimal with respect to the Singleton-type bound presented by Prakash \emph{et al.} New upper bounds on the length of such optimal codes are derived. The new bounds both improve and generalize previously known bounds. We also talk about optimal codes, whose length is order-optimal when compared with the new upper bounds. The length of the codes is super-linear in the alphabet size.
[Full version]

Floater–Hormann interpolants constitute a family of barycentric rational interpolants based on the blend of local polynomial interpolants of degree d that have recently proved to be a viable alternative to more classical interpolation methods, such as polynomials and splines, especially in the equispaced setting. In the first part of this seminar we focus on the approximation of the derivatives of a function f with classical Floater–Hormann interpolants and show how their k-th derivative converge to ...
[Full version]

Shape correspondence is a fundamental task in shape analysis, with a variety of applications in computer graphics and computer vision. Example applications include statistical shape interpolation, texture, segmentation and deformation transfer, and application of deep learning to 3D shapes. Formulating the desired geometric properties of the correspondence is a challenging task, as the quality of the result can be easily evaluated by a human observer, yet is challenging to quantify. Additionally, the desired properties of ...
[Full version]

Shape correspondence is a fundamental task in shape analysis, with a variety of applications in computer graphics and computer vision. Example applications include statistical shape interpolation, texture, segmentation and deformation transfer, and application of deep learning to 3D shapes. Formulating the desired geometric properties of the correspondence is a challenging task, as the quality of the result can be easily evaluated by a human observer, yet is challenging to quantify. Additionally, the desired properties of ...
[Full version]

We show that Folded Reed-Solomon codes achieve list decoding capacity with constant list sizes, independent of the block length. Prior work yielded list sizes that are polynomial in the block length, and relied on elaborate subspace evasive machinery to reduce the list sizes to constant. We further show that multiplicity codes exhibit similar behavior, and use this to obtain capacity achieving locally list decodable codes with query complexity significantly lower than was known before. Based ...
[Full version]

With their introduction in 2009, cryptocurrencies brought the promise of a global decentralized payment system. However, protocol limitations prevent them from managing the required throughput. Previous work overcame protocol limitations and used low-degree sharding to allow for parallelization. All these systems rely on efficient validation of the transactions, in each shard or globally. We present Ostraka, an architecture for arbitrary scaling of blockchain nodes that does not affect the security properties of the underlying protocol. ...
[Full version]

We will consider the SITA server farm scheduling policies which were introduced and studied by Harchol-Balter and her collaborators. In particular we will discuss a recent INFOCOM17 paper of Doncel, Aalto and Ayesta and relate one of the tables there to the work of Riemann on the functional equation of the zeta function. We will also consider certain disk drive scheduling problems (travelling salesman on a disk drive) and then relate it to space-time geometry ...
[Full version]

Faculty and Graduate students are invited to a workshop on “Machine Learning on the AWS Cloud” on Monday, May 13, 2019, 15:00-18:30, in room 337, CS Taub Building. More details and program in the attached poster and document. Please RSVP by email to: awstomer at amazon.com
[Full version]

Artificial illumination plays a vital role in human civilization. In computer vision, artificial light is extensively used to recover 3D shape, reflectance, and further scene information. However, in most computer vision applications using artificial light, some additional structure is added to the illumination to facilitate the task at hand. In this work, we show that the ubiquitous alternating current (AC) lights already have a valuable inherent structure that stems from bulb flicker. By passively sensing ...
[Full version]

Mobile sensing is an emerging research area with many applications, including environmental monitoring and medical imaging. For example, tomography, which is used extensively in medical imaging, can be re-formulated by means of sampling along trajectories in the sinogram domain. We propose a model that accounts for finite-energy functions and present continuous-domain analysis accordingly. We then consider the problem of reconstructing CT scans within the mobile sensing framework. We initially focused on 1D signals and found ...
[Full version]

Consider the following hat guessing game: nn players are placed on nn vertices of a graph, each wearing a hat whose color is arbitrarily chosen from a set of qq possible colors. Each player can see the hat colors of his neighbors, but not his own hat color. All of the players are asked to guess their own hat colors simultaneously, according to a predetermined guessing strategy and the hat colors they see, where no ...
[Full version]

We are happy to invite your to the fourth of series of meetings on career and job seeking which will be held at CS, on Wednesday, May 1, at 12:30, in Taub Classroom 7, CS Taub Building. Mr. David Gross, Program Engineer in Facebook Israel will give a talk on how to prepare and succeed in a technical interview: "Crush Your Coding Interview". Please pre-register. See you there!
[Full version]

We study the communication complexity of welfare maximization in combinatorial auctions with m items and two subadditive bidders. A 12-approximation can be guaranteed by a trivial randomized protocol with zero communication, or a trivial deterministic protocol with O(1) communication. We show that outperforming these trivial protocols requires exponential communication, settling an open question of [DobzinskiNS10, Feige09]. Specifically, we show that any (randomized) protocol guaranteeing a (12+6log2m)-approximation requires communication exponential in m. This is tight even ...
[Full version]

In recent years, cryptocurrencies, and blockchain as their underlying technology have emerged. They act as a decentralized ledger, i.e., an append-only log, where payments are recorded. One of the key limitations of almost all blockchains is the overall throughput, e.g., in Bitcoin, the first cryptocurrency, the overall throughput is about 7 transactions per second (TPS), whereas VISA’s maximum throughput is between three to four orders of magnitude higher. One of the main reasons this problem ...
[Full version]

Automated agents that interact proficiently with people can be useful in supporting, training or replacing people in complex tasks. The inclusion of people presents novel problems for the design of automated agents’ strategies. People do not necessarily adhere to the optimal, monolithic strategies that can be derived analytically. Their behavior is affected by a multitude of social and psychological factors. In this talk I will show how combining machine learning techniques for human modelling, human ...
[Full version]

A deep learning approach to numerically approximate the solution to the Eikonal equation is introduced. The proposed method is built on the fast marching scheme which comprises of two components: a local numerical solver and an update scheme. We replace the formulaic local numerical solver with a trained neural network to provide highly accurate estimates of local distances for a variety of different geometries and sampling conditions. Our learning approach generalizes not only to flat ...
[Full version]

We propose a controllable geometric flow that decomposes the interior volume of a triangular mesh into a collection of encapsulating layers, which we denote by a generalized foliation. For star-like genus zero surfaces we show that our formulation leads to a foliation of the volume with leaves that are closed genus zero surfaces, where the inner most leaves are spherical. Our method is based on the three-dimensional Hele-Shaw free-surface injection flow, which is applied to ...
[Full version]

We propose a controllable geometric flow that decomposes the interior volume of a triangular mesh into a collection of encapsulating layers, which we denote by a generalized foliation. For star-like genus zero surfaces we show that our formulation leads to a foliation of the volume with leaves that are closed genus zero surfaces, where the inner most leaves are spherical. Our method is based on the three-dimensional Hele-Shaw free-surface injection flow, which is applied to ...
[Full version]

Many intractable computational problems on graphs admit tractable algorithms when applied to trees or forests. In such cases, a tree decomposition of the input graph often facilitates an effective solution. Therefore, tree decompositions are critical for graph problems in the fields of databases, game theory, statistical analysis, bioinformatics and more. The goal of this research is to build an efficient tool for enumerating (producing) a large space of tree decompositions for solvers of graph problems ...
[Full version]

The 2019 open day invites outstanding undergraduates from all universities to come to the Technion and learn about the Computer Science and Department, to meet faculty and graduate students and to hear a fascinating talk by Dr. Tal Cohen, Director of Program Development at Google: "Do Advanced Degrees Indeed Advance?" The event will be held on Wednesday, April 17, 2019, between 12:00, in CS Taub Building, room 337 (3rd floor). The program will include review on ...
[Full version]

Modern machine learning models exhibit super-human accuracy on tasks from image classification to natural-language processing, but accuracy does not tell the entire story of what these models have learned. Does a model memorize and leak its training data? Does it contain hidden backdoor functionality? In this talk, I will explain why common metrics of model quality may hide potential security and privacy vulnerabilities, and outline recent results and open problems at the junction of machine ...
[Full version]

Question Answering (QA) is one of the core tasks in natural language understanding. This task requires the ability to process and understand natural language questions and documents, as well as to extract the answers to these questions. There are several settings for the QA task, which provide varying levels of difficulty. In the most simple setting, a question is paired with a context which is guaranteed to contain sufficient evidence to answer the question. A ...
[Full version]

This talk presents some of the geometrical aspects involved in treating irregular heart beat rhythm (Arrythmia) using Carto 3 System. Carto 3 is a product of Biosense-Webster, a global leader in the science of diagnosing and treating heart rhythm disorders. CARTO 3 System enables accurate visualization of multiple catheters in a patient’s heart and pinpoints exact location/orientation of a catheter. During arrythmia procedure, a 3D electro-anatomical reconstruction of the heart is built and color coded ...
[Full version]

We prove a new connection between gentle measurement (where one wants to measure n quantum states, in a way that damages the states only by a little) and differential privacy (where one wants to query a database about n users, in a way that reveals only a little about any individual user). The connection is bidirectional, though with loss of parameters in going from DP to gentle measurement. Exploiting this connection, we present a new ...
[Full version]

The talk will survey new developments that use cryptographic tools in order to use a quantum computer even if it is operated by an untrusted party. Applications include: proofs of quantumness, certifiable randomness, privately outsourcing quantum computation (via quantum fully homomorphic encryption) and delegating quantum computation.
[Full version]

Information Extraction (IE) is the task of extracting structured information from textual data. We explore a programming paradigm that is supported by several IE systems where relations extracted by atomic extractors undergo a relational manipulation. In our efforts toward achieving a better understanding of IE systems, we study queries in the framework of document spanners that was introduced by Fagin et al. A spanner is a function that extracts from a document (string) a relation ...
[Full version]

Partial differential equations (PDEs) are indispensable for modeling many physical phenomena and also commonly used for solving image processing tasks. In the latter area, PDE-based approaches interpret image data as discretizations of multivariate functions and the output of image processing algorithms as solutions to certain PDEs. Posing image processing problems in the infinite dimensional setting provides powerful tools for their analysis and solution. Over the last few decades, the reinterpretation of classical image processing problems ...
[Full version]

Machine learning algorithms have difficulties to generalize over a small set of examples. Humans can perform such a task by exploiting vast amount of background knowledge they possess. One method for enhancing learning algorithms with external knowledge is through feature generation. We introduce a new algorithm for generating features based on a collection of auxiliary datasets. We assume that, in addition to the training set, we have access to a additional datasets. We do not ...
[Full version]

e-lections 2019 event, organized by Prof. Eli Biham, Orr Dunkelman, Yossi Oren and the Technion Hiroshi Fujiwara Cyber Security Research Center, will be held on Sunday, April 7, 2019 at the Technion, The event will present lectures related to electronic elections and to the effect of the cyber world on elections and will include lectures on the Brazilian and Estonian computerized elections, as well as influence on Israeli elections. The lectures will be given in ...
[Full version]

There are two well studied security notions for public key encryption schemes: security against passive adversaries and security against active adversaries. In this talk, we will discuss a new generic transformation that can upgrade the security of an encryption scheme from passive security to active security. For this transformation, we use a special family of pseudorandom generators (PRGs) called hinting PRGs. I will first introduce these special PRGs, followed by an outline of our transformation. ...
[Full version]

Model checking is an automatic verification method that accepts a system model and a specification, and checks whether the model satisfies the specification. CTL is a branching temporal logic suitable for specifying behaviors of both software and hardware systems. In this work, we present a novel approach that combines on-the-fly verification with abstraction in order to obtain an efficient model checking. On-the-fly verification ensures that only parts that are needed for determining the satisfaction of ...
[Full version]

In this talk I will tell you how analyzing economic markets where agents have cognitive biases has led to better understanding of the communication complexity of local search procedures. We begin the talk with studying combinatorial auctions with bidders that exhibit endowment effect. In most of the previous work on cognitive biases in algorithmic game theory (e.g., [Kleinberg and Oren, EC'14] and its follow-ups) the focus was on analyzing the implications and mitigating their negative ...
[Full version]

We consider the Graph Balancing problem, where we are given an undirected multigraph with edge weights and orientation costs. The goal is to find an orientation of the edges that minimizes the worst weighted incoming degree of a vertex (also known as makespan) while minimizing the total orientation cost. Graph Balancing is an interesting special case of the well-known Generalized Assignment Problem (GAP), which is perhaps one of the most famous and extensively studied problems ...
[Full version]

The hexahedral meshing problem is the volumetric analog of the quad meshing problem, with analogous applications in finite element modeling. One might expect that frame field–based methods, which have proven effective in quad meshing, would extend naturally to the volumetric case. Unfortunately, several problems arise in attempting this extension. The geometry of frames becomes significantly more complicated, making optimization over frames more challenging. Moreover, field and mesh topology is far more complicated in 3D due ...
[Full version]

We study the Excluded Grid Theorem, a fundamental structural result in graph theory, that was proved by Robertson and Seymour in their seminal work on graph minors. The theorem states that there is a function f, such that for every integer g > 0, every graph of treewidth at least f(g) contains the (g×g)-grid as a minor. For every integer g>0, let f(g) be the smallest value for which the theorem holds. Establishing tight bounds ...
[Full version]

Unravelling underlying complex structures from limited resolution measurements is a known problem arising in many scientific disciplines. We study a stochastic dynamical model with a multiplicative noise. It consists of a stochastic differential equation living on a graph, similar to approaches used in population dynamics or directed polymers in random media. This model has numerous applications, such as population growth, economic growth, and optimal control. We present a new application of this model in the ...
[Full version]

Blockchain protocols are notoriously bad at scaling. Most protocols do work reasonably well when used in a small scale. Yet, as the network size grows, resources are added and overall computing power increases, most protocols simply do not scale. This is largely due to the fact that the entire process is fully replicated by all participants. Surpassing some desired redundancy threshold, this offers no advantage. We tackle the issue by introducing a formal general framework ...
[Full version]

Fictitious domain methods are an elegant way to circumvent a tedious discretization process, mostly meshing. The core idea is to embedded the geometrically complex model into a fictitious domain, which has zero stiffness assigned. The complexity of the model is then recovered on the integration level, where the only information required from the model is an unambiguous, reliable point membership classification (PMC) on the integration points. This point membership classification can be easily carried out ...
[Full version]

We consider fault-tolerant boolean formulas in which the output of a faulty gate is short-circuited to one of the gate’s inputs. A recent result by Kalai et al. [FOCS 2012] converts any boolean formula into a resilient formula of polynomial size that works correctly if less than a fraction 1/6 of the gates (on every input-to-output path) are faulty. We improve the result of Kalai et al., and show how to efficiently fortify any boolean ...
[Full version]

You are invited to the final stage of the 2019 Best Project Contest. The finalist teams will present and talk about their projects. The event will take place on Wednesday, March 20, 2019, between 12:00-14:30, at the lobby of CS Taub Building. You are all invited to cheer, support and to inspect the most outstanding CS projects.
[Full version]

We present a novel method for computing correspondences between subdivision surfaces, represented as functional maps. We leverage two main advantages of the hierarchical structure: (1) the spectral basis functions, which are multi-resolution by construction, can be succinctly represented on a subdivision surface, and (2) the pointwise map reconstruction can be done hierarchically. Using these properties, we design a computational pipeline for finding correspondences between high resolution meshes which is much faster than existing approaches. We ...
[Full version]

Puzzle solving is a difficult problem in its own right, even when the pieces are all square and build up a natural image. But what if these ideal conditions do not hold? One such application domain is archaeology, where restoring an artifact from its fragments is highly important. From the point of view of computer vision, archaeological puzzle solving is very challenging, due to three additional difficulties: the fragments are of general shape; they are ...
[Full version]

Today we are literally surrounded by software. Almost all products and services we use on a daily basis involve some piece of software. The abundance of software has made our lives easier. Unfortunately it also made us susceptible than ever to software vulnerabilities and exploitation. A big part of the effort to secure software is often carried out by researchers external to the team that developed the software. A crucial tool in their tool bag ...
[Full version]

Billions of dollars a year are spent to develop a new drug. The first step in the drug development process is drug discovery, where in potential molecules are studied before leads are selected and moved to clinical trials. Extensive research is performed to identify studies that might be relevant to the potential molecule or its substructures. As the molecule is novel, no respective research has been published about it, and therefore identifying relevant papers about ...
[Full version]

Neural Architecture Search (NAS) aims to facilitate the design of deep networks for a given task. This is part of a larger trend – automated machine learning (AutoML) – that promises to solve or at least alleviate the scarcity of ML experts needed to design custom architectures. We are working on a differentiable search method unlike conventional approaches of applying evolution or reinforcement learning methods. Recent works show that it is possible to reduce a ...
[Full version]

In this lecture we will introduce a few quantum complexity classes (BQP, QMA, QIP, and QPIP), the Deutsch-Jozsa algorithm, and Simon's algorithm; we will finally (very briefly) discuss Shor's algorithm for factoring large numbers.
[Full version]

We have witnessed worldwide network outages after large-scale natural disasters, such as earthquakes, hurricane, and terrorist attacks. In the talk, I will discuss how to extend the widely implemented single link failure models for regional failures. I will mainly focus on the related theoretical problems in the domain of combinatorics and computational geometry. First, I will show that operators need to prepare their network only against a small number of failure-states. Second, I introduce a ...
[Full version]

CYBERDAY 2019 event, organized by Prof. Eli Biham, Prof. Sara Bitan, and the Technion Hiroshi Fujiwara Cyber Security Research Center, will be held on Tuesday, March 12, 2019 at the Technion, Keynote speakers: Prof. Bart Preneel (KU Leuven, Belgium) Prof. Adi Shamir (The Weizmann Institute of Science). Most of the lectures will be given in Hebrew and a poster session will take place at the break: graduate students and all who are interested to present ...
[Full version]

In this talk, we will discuss problems associated with synchronizing specific events in space and time (X-coherence), fact-checking, and mining persons, objects and contents of interest from various and heterogeneous sources including — but not limited to — the internet, social media and surveillance imagery. For that, we seek to harness information from various media sources and synchronize the multiple textual and visual information pieces around the position of an event or object as well ...
[Full version]

In this lecture we will introduce the quantum bit (qubit), the basic rules of quantum information/computing, the no-cloning principle, quantum gates for universal quantum computing; finally we will see two protocols, one for quantum teleportation, and one for quantum key distribution.
[Full version]

Nowadays, as computer networks continuously increase in size and speed, making efficient use of compute and network resources becomes exceptionally challenging. At the heart of this challenge are scheduling and load balancing techniques that are expected to optimize resource usage and, in turn, minimize costs. In this talk we shall address two such emerging network challenges. In the first part of the talk we shall present dRMT (disaggregated Reconfigurable Match-Action Table), a new architecture for ...
[Full version]

In this thesis, we show how geometry, and specifically parametric freeform geometry, can be used to solve manufacturing related problems. This thesis deals with both additive manufacturing (specifically 3D printing) and the more traditional (subtractive manufacturing) machining process. The first topic handled is the representation and manufacturing of functionally graded material (FGM) objects in 3D printing. Specifically, this thesis presents the use of parametric (trimmed) volumes to manufacture FGM objects. Using the underlying parametrization offered ...
[Full version]

Named-entity Recognition (NER) is an important task in the NLP field, and is widely used to solve many challenges. However, in many scenarios, not all of the entities are explicitly mentioned in the text. Sometimes they could be inferred from the context or from other indicative words. Consider the following sentence: CMA can easily hydrolyze into free acetic acid. Although water is not mentioned explicitly, one can infer that H2O is an entity involved in ...
[Full version]

In bidirectional heuristic search the search proceeds from both the start and the goal simultaneously.The state of the art in bidirectional search has changed significantly a very short time period; we now can answer questions about unidirectional and bidirectional search that until very recently we were unable to answer. This talk is designed to provide an accessible overview of the recent research in bidirectional search in the context of the broader efforts over the last ...
[Full version]

We propose a principled approach to integrating GPU memory with an OS page cache. We design GAIA, a weakly-consistent page cache that spans CPU and GPU memories. GAIA enables the standard mmap system call to map files into the GPU address space, thereby enabling data-dependent GPU accesses to large files and efficient write-sharing between the CPU and GPUs. Under the hood, GAIA (1) integrates lazy release consistency among physical memories into the OS page cache ...
[Full version]

Rapid advances in data-driven applications over recent years have intensified the need for efficient mechanisms capable of real-time monitoring and detecting arbitrarily complex patterns in massive data streams. Complex event processing (CEP) is a prominent technology widely employed for performing this task in many areas, including online finance, security monitoring, credit card fraud detection, and IoT (Internet of Things) technologies. An increasingly active and rapidly developing area of academic research, CEP functionality is also provided ...
[Full version]

Inverse problems appear in many applications, such as image deblurring, inpainting and super-resolution. The common approach to address them is to design a specific algorithm (or recently - a deep neural network) for each problem. The Plug-and-Play (P&P) framework, which has been recently introduced, allows solving general inverse problems by leveraging the impressive capabilities of existing denoising algorithms. While this fresh strategy has found many applications, a burdensome parameter tuning is often required in order ...
[Full version]

The wide usage of smartphones encourages users to ubiquitously and constantly interact with web applications like microblogs and search engines. Examples of atomic interactions, referred as micro-actions, include posing a query, sending a textual message, sharing a photo etc. Analysis of micro-actions is beneficial for understanding the context of a micro-action and can be used for improving existing services and offering new ones. For example, by detecting whether a search is performed from a location ...
[Full version]

In the Maximum Carpool Matching problem we seek for the best way a group of people can share their ride based on their personal preferences (music, smoking, etc...). Submodularity is a property of set functions. The problem of maximizing a submodular function appears in many applications such as viral marketing, information gathering, image segmentation, document summarization, and speeding up satisfiability solvers. Both the Maximum Carpool Matching and the Submodular Function Maximization problems are NP-hard, and ...
[Full version]

Patch matching and local treatment of images has proven to be a successful strategy for problems such as image denoising, inpainting, super-resolution, image editing and the list could go on and on. Naturally, the size of the patch is a crucial parameter of the patch matching algorithm. Practically, working with large patches often results in a deterioration of reconstruction performance. This is due the curse of dimensionality -- an increase in the patch-size requires an ...
[Full version]

Feed-forward CNNs trained for image transformation problems rely on loss functions that measure the similarity between the generated image and a target image. Most of the common loss functions assume that these images are spatially aligned and compare pixels at corresponding locations. However, for many tasks, aligned training pairs of images will not be available. We present an alternative loss function that does not require alignment, thus providing an effective and simple solution for a ...
[Full version]

The mathematical study of social and information networks has historically centered around generative models for such networks (preferential attachment, the Chung-Lu random graph model, Kronecker graphs, etc.). This talk proposes distribution-free models of social and information networks - novel classes of graphs that capture all plausible such networks. Our models are motivated by triadic closure, the property that vertices with one or more mutual neighbors tend to also be neighbors; this is one of the ...
[Full version]

Rich engagement data can shed light on how people interact with online content and how such interactions may be determined by the content of the page. In this work, we investigate a specific type of interaction, backtracking, which refers to the action of scrolling back in a browser while reading an online news article. We leverage a dataset of close to 700K instances of more than 15K readers interacting with online news articles, in order ...
[Full version]

Recent years have seen the emergence of new kinds of software including deep learning, programmable computer networks, and blockchains. Unfortunately, these systems have been shown to suffer from critical safety and security errors, affecting their wider adoption. The goal of my research is to develop new automated program verification and synthesis techniques which ensure safety and reliability of these systems. In this talk, I will start by introducing AI2, the first automated verifier for neural ...
[Full version]

The chest X-ray scan is by far the most commonly performed radiological examination for screening and diagnosis of many cardiac and pulmonary diseases. It is also one of the hardest to interpret, with disagreement rating of around 30% even for experienced radiologists. At Zebra, we have access to millions of X-ray scans, as well as their accompanied anonymized textual reports written by hospital radiologists. Can this data be used to teach an algorithm to identify ...
[Full version]

With the increase in malicious activity over the Internet, it has become extremely important to build tools for automatic detection of such activity. There have been attempts to use machine learning to detect network attacks, but the difficulty in obtaining positive (attack) examples, led to using one-class methods for anomaly detection. In this work we present a novel framework for using multiclass learning to induce an attack detector that identifies attacks at run time. We ...
[Full version]

Search systems are often focused on providing relevant results for the "now", assuming both corpora and user needs that focus on the present. However, many corpora today reflect significant longitudinal collections ranging from 20 years of the Web to hundreds of years of digitized newspapers and books. Understanding the temporal intent of the user and retrieving the most relevant historical content has become a significant challenge. Common search features, such as query expansion, leverage the ...
[Full version]

The round complexity of zero-knowledge protocols is a long-standing open question, yet to be settled under standard assumptions. So far, the question has appeared equally challenging for relaxations such as weak zero-knowledge and witness hiding. Protocols satisfying these relaxed notions under standard assumptions have at least four messages, just like full-fledged zero knowledge. The difficulty in improving round complexity stems from a fundamental barrier: none of these notions can be achieved in three messages via ...
[Full version]

The CS Computer Communication Lab (LCCN) invites you to an exposure evening to communication networks - to get to know the laboratory activity, the research and the projects it offers and the team that leads them, to listen to lectures and take part in discussions on the matter, as described in the attached ad. The event will take place on Tuesday, January 22, 2019 at 18:30, in room 337 (3rd floor). Students to all degrees ...
[Full version]

CS Labs: Systems and Software Development Laboratory (SSDL), Cyber and Computer Security Laboratory (CYBER), The Laboratory for Computer Communication and Networking (LCCN) invite you to visit the Yearly Project Fair on IoT, Android, Ransomware and Networks. The fair will be held on Tuesday, January 22, 2019, starting at 120, in the transparent hall of Beit Hastudent. You are all invtied! The presenting projects.
[Full version]

Neural Networks have been shown to be remarkably resistant to label noise. This means that you can train a network using a data set that contains a significant fraction of wrongly-labelled samples, and the network will still be able to accurately predict the label for a previously unseen sample. Our results show that a main factor in explaining this resistance is that networks learn from a local group of training samples, similarly to a K-Nearest ...
[Full version]

The question whether one can recover the shape of a geometric object from its Laplacian spectrum (‘hear the shape of the drum’) is a classical problem in spectral geometry with a broad range of implications and applications. While theoretically the answer to this question is negative (there exist examples of iso-spectral but non-isometric manifolds) little is known about the practical possibility of using the spectrum for shape reconstruction and optimization. In this talk, I will ...
[Full version]

The rapid increase in the popularity of cryptocurrencies brought with it not only questions regarding the quality of the technology itself, but also the economic phenomena surrounding them. Bitcoin in particular has proven itself to be a very volatile commodity~[Baek and Elbeck, Applied Economics 2015], and a poor currency, according to the definition of money [Kocherlakota, Journal of Economic Theory]. This has given rise to numerous alternative cryptocurrency models, which set about to improve upon ...
[Full version]

Given a set of points in a metric space, the Closest Pair problem asks to find a pair of distinct points in the set with the smallest distance. In this talk, we address the fine-grained complexity of this problem which has been of recent interest. At the heart of all our proofs is the construction of a dense bipartite graph with special embedding properties and are inspired by the construction of locally dense codes. The ...
[Full version]

Stream frequency measurements are fundamental in many data stream applications such as financial data trackers, intrusion-detection systems, and network monitoring. Typically, recent data items are more relevant than old ones, a notion we can capture through a sliding window abstraction. This paper considers a generalized sliding window model that supports stream frequency queries over an interval given at query time. This enables drill-down queries, in which we can examine the behavior of the system in ...
[Full version]

How can we capture effectively the information expressed in multiple texts? How can we allow people, as well as computer applications, to easily explore it? The current semantic NLP pipeline typically ends at the single sentence or text level, putting the burden on applications to consolidate and present related information across multiple texts. Further, semantic representations, which may provide the basis for text consolidation, are often based on non-trivial schemata which require expert annotation, making ...
[Full version]

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

A model which mimics the programming operation of memory cells was first presented by Bunte and Lapidoth. Under this paradigm we assume that cells are programmed sequentially and individually. The programming process is modeled as transmission over a channel, while it is possible to read the cell state in order to determine its programming success, and in case of programming failure, to reprogram the cell again. Reprogramming a cell can reduce the bit error rate, ...
[Full version]

A central object in optimal stopping theory is the single-choice prophet inequality for independent, identically distributed random variables: Given a sequence of random variables X_1,…,X_n drawn independently from a distribution F, the goal is to choose a stopping time τ so as to maximize α such that for all distributions F we have E[X_τ] ≥ α⋅E[max_t X_t]. What makes this problem challenging is that the decision whether τ = t may only depend on the ...
[Full version]

Widely recognized as the inventor of digital cash, David is also known for other fundamental innovations in cryptography, including privacy and secure election technology. David is currently leading the Elixxir blockchain project, a private, scalable platform capable of processing a high volume of payments, messages and file transfers. David will discuss Elixxir's breakthrough mixed network and multi-party computation architecture and how the technology addresses the current scaling and privacy challenges limiting blockchain adoption.
[Full version]

As technology becomes highly integrated into nearly every aspect of life, and in many cases is indispensable, maintaining relative hand functionality is crucial. A transradial amputation is a partial amputation of the arm below the elbow that can greatly impede one’s ability to perform every day tasks at home and at work. With the growing dependency on hand functionality and gesture formation to keep up with technological use, the disparity in functional ability between those ...
[Full version]

We tackle the problem of multiple image alignment and 3D reconstruction under extreme noise. Photographs acquired in scenes with large intensity variations or in scattering media have significant spatial variance in signal-to-noise-ratio (SNR), where in some image regions it is significantly below the level current methods cope with. Modern alignment schemes, based on similarity measures, feature matching and optical flow are often pairwise, or assume global alignment. Nevertheless, under extreme noise, the alignment success sharply ...
[Full version]

Can we program computers in our native tongue? This idea, termed natural language programming (NLPRO), has attracted attention almost since the inception of computers themselves. From the point of view of software engineering (SE), efforts to program in natural language (NL) have relied thus far on controlled natural languages (CNL) -- small unambiguous fragments of English with restricted grammars and limited expressivity. Is it possible to replace these CNLs with truly natural, human language? From ...
[Full version]

We consider the problem of anomaly detection in images, and present a new detection technique. Given a sample of images, all known to belong to a ``normal'' class (e.g., dogs), we show how to train a deep neural model that can detect out-of-distribution images (i.e., non-dog objects). The main idea behind our scheme is to train a multi-class model to discriminate between dozens of geometric transformations applied on all the given images. The auxiliary expertise ...
[Full version]

In the era of “Big Code”, research is being conducted into automating the understanding of computer programs. Most of the current works borrow techniques from natural language processing and deep learning, which have been successful recently, attempting to process the code directly or using syntactic representations (e.g., ASTs and AST paths). However, to comprehend program semantics robustly, structural features of code have to be taken into account as well, including function calls, branching, and interchangeable ...
[Full version]

Private information retrieval (PIR) is a technique of retrieving data from servers in a distributed storage system (DSS), without revealing the identity of the data to the servers. A scheme that achieves the PIR property is referred to as a PIR protocol and is characterized by its PIR rate, which the ratio of downloaded data and the size of the requested data. The supremum of PIR rates achieved by any PIR protocol in a DSS ...
[Full version]

It is no secret that online companies, hospitals, credit-card companies and governments hold massive datasets composed of our sensitive personal details. Information from such datasets is often released using some privacy preserving heuristics, which have been repeatedly shown to fail. That is why in recent years the notion of differential privacy has been gaining much attention, as an approach for conducting data-analysis that adheres to a strong and mathematically rigorous notion of privacy. Indeed, many ...
[Full version]

Understanding deep learning calls for addressing three fundamental questions: expressiveness, optimization and generalization. Expressiveness refers to the ability of compactly sized deep neural networks to represent functions capable of solving real-world problems. Optimization concerns the effectiveness of simple gradient-based algorithms in solving non-convex neural network training programs. Generalization treats the phenomenon of deep learning models not overfitting despite having much more parameters than examples to learn from. This talk will describe a series of works ...
[Full version]

n the function-inversion problem, an algorithm gets black-box access to a function $f:[N] \to [N]$ and takes as input a point $y \in [N]$, along with $S$ bits of auxiliary information about $f$. After running for time $T$, the algorithm must output an $x \in [N]$ such that $f(x) = y$, if one exists. This problem, first studied by Hellman (1980), has manifold applications to cryptanalysis. Hellman’s algorithm for this problem achieves the time-space tradeoff ...
[Full version]

Hardware security of integrated circuits (ICs) is a significant concern in many emerging market segments, such as intelligent transportation, innovative health care, sophisticated security systems, and smart energy applications. In particular, modern side-channel attacks that exploit information leaked by physical IC implementation pose a serious threat to the security of secret data on-chip. In this talk, a theoretical framework will be presented for actively detecting power analysis attacks (PAAs) – one of the commonly used ...
[Full version]

Programming languages are a tool for human thought, expression, and work yet they are principally designed using mathematical and engineering techniques. In this talk, I will describe how our group has applied human-centered design techniques --- interviews, participatory design exercises, and qualitative analysis of developer forums --- in the design of three research programming systems (Plaid, Glacier, and Obsidian). I will speak frankly about the strengths and weaknesses of these approaches and discuss speculative new ...
[Full version]