אירועים והרצאות בפקולטה למדעי המחשב ע"ש הנרי ומרילין טאוב

We consider the 0-Extension problem, where we are given an undirected graph G=(V,E) equipped with non-negative edge weights w: E→ℝ+, a collection T={t1,…,tk}⊆V of k special vertices called terminals, and a semi-metric D over T. The goal is to assign every non-terminal vertex to a terminal while minimizing the sum over all edges of the weight of the edge multiplied by the distance in D between the terminals to which the endpoints of the edge are assigned. 0-Extension admits two know...
[ לנוסח המלא ]

Thin structures are ubiquitous, such as ropes, cables, tree branches, wire arts, and wire-frame furniture. It is well known that 3D thin structures are challenging to reconstruct due to their narrow width and lack of distinct features. However, their reconstruction has received relatively little research attention. Clearly, the detection and reconstruction of these thin objects have many applications in computer graphics, computer vision, and robotics. In this talk I will present our recent work...
[ לנוסח המלא ]

Our research is focused on two distributed models that allow global communication. The first one is the well-known Congested Clique model and another is the recently introduced Hybrid network model. In this talk, I will present three scheduling algorithms for the Congested Clique model, which we developed during my M.Sc studies. These algorithms run a set of distributed jobs in near-optimal time in a black-box manner, without a priori knowledge of their communication pattern. ...
[ לנוסח המלא ]

Correlation Clustering is an elegant model where given a graph with edges labeled + or −, the goal is to produce a clustering that agrees the most with the labels: + edges should reside within clusters and − edges should cross between clusters. In this work we study the MaxCorr objective, aiming to find a clustering that maximizes the difference between edges classified correctly and incorrectly. We focus on the case of bipartite graphs and present an improved approximation of 0.254, improvi...
[ לנוסח המלא ]

Storage processors are a new type of acceleration device that offloads centerpiece storage stack functions from legacy software to high-performance hardware. The legacy software paradigms waste flash capacity and performance, pushing the users to deploy expensive and overprovisioned SSDs. For example, the high write amplification induced by inefficient software limits the applicability of the next-generation low-cost flash technologies like QLC and PLC, as well as ZNS SSDs. Storage processors ar...
[ לנוסח המלא ]

3D computer vision has significantly advanced over the past several decades, with modern algorithms successfully reconstructing entire urban cities. However, many questions remain unexplored, as geometric reasoning alone cannot fully infer the connections among images capturing different parts of the scene or semantic relationships between images captured at distant geographic locations. In this talk, I will present an ongoing line of research that leverages powerful deep networks to addres...
[ לנוסח המלא ]

We introduce a new deep learning method for point cloud comparison. Our approach, named Deep Point Cloud Distance (DPDist), measures the distance between the points in one cloud and the estimated surface from which the other point cloud is sampled. The surface is estimated locally and efficiently using the 3D modified Fisher vector representation. The local representation reduces the complexity of the surface, enabling efficient and effective learning, which generalizes well between object categ...
[ לנוסח המלא ]

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...
[ לנוסח המלא ]

Evolving applications such as autonomous vehicle teleoperation call for high-quality and low latency video transmission. Existing technology allows high-quality video to be transmitted over several parallel cellular channels, with suitable rate control. Leveraging these systems to vehicles in motion involves further challenges due to the higher variability of the cellular channels. Major players in these technologies are the Israeli-based company LiveU and its spinoff DriveU. Here, we present a...
[ לנוסח המלא ]

Programmable switches are advanced network switches that can be programmed to perform certain actions besides routing packets. Programmable switches enable combining the logic and flexibility of software with the speed and efficiency of hardware and thus enjoy the benefits of both. Interstingly, programmable switches are also 2-3 orders of magnitude more power to compute efficient than GPUs and FPGAs. Alas, they offer a very restricted and non-intuitive programming model. Sorting is a fundame...
[ לנוסח המלא ]

We study the social efficiency of contracts as economic mechanisms when multiple principals simultaneously manage a common agent. We consider an incomplete-information setting: The agent chooses an unobservable action that induces both a cost for the agent and an expected value for each principal. The sum of these terms is referred to as the resulting social welfare. The agent’s choice is incentivized by payment schemes (“contracts”) set forth by the principals, who have different private ...
[ לנוסח המלא ]

We study the problem of maximizing a monotone submodular function subject to a Multiple Knapsack constraint (SMKP). The input is a set I of items, each associated with a non-negative weight, and a set of bins, each having a capacity. Also, we are given a submodular, monotone and non-negative function f over subsets of the items. The objective is to find a subset of items A⊆I and a packing of the items in the bins, such that f(A) is maximized. SMKP is a natural extension of both Multiple Knapsa...
[ לנוסח המלא ]

Almost two centuries ago Pierre-Joseph Proudhon proposed social contracts -- voluntary agreements among free people -- as a foundation from which an egalitarian and just society can emerge. A digital social contract is the novel incarnation of this concept for the digital age. Digital social contracts are: Conceptually, voluntary agreements among genuinely-identified people, specified, undertaken, and fulfilled in the digital realm. Mathematically, a novel computational model specifying concurre...
[ לנוסח המלא ]

Recent advances in material science and digital fabrication provide promising opportunities for product design, mechanical and biomedical engineering, robotics, architecture, art, and science. Engineered materials and personalized fabrication are revolutionizing manufacturing culture and having a significant impact on various scientific and industrial works. As new fabrication technologies emerge, effective computational tools are needed to fully exploit the potential of digital fabrication. In ...
[ לנוסח המלא ]

The ability to promptly and efficiently detect arbitrarily complex patterns in massive real-time data streams is a crucial requirement for a wide range of modern applications. The ever-growing scale of these applications and the sophistication of the patterns involved makes it imperative to employ advanced solutions that can optimize pattern detection. One of the most prominent and well-established ways to achieve the above goal is to apply complex event processing (CEP) in a parallel manner, us...
[ לנוסח המלא ]

Bayesian persuasion, as introduced by Kamenica and Gentzkow in 2011, is the study of information sharing policies among strategic agents. A prime example is signaling in online ad auctions: what information should a platform signal to an advertiser regarding a user when selling the opportunity to advertise to her? Practical considerations such as preventing discrimination, protecting privacy or acknowledging limited attention of the information receiver impose constraints on information sharing....
[ לנוסח המלא ]

A streaming algorithm is given a long sequence of items and seeks to compute or approximate some function of this sequence using a small amount of memory. A body of work has been developed over the last two decades, resulting in optimal streaming algorithms for a wide range of problems. While randomized streaming algorithms are well-studied, the vast majority of them are defined and analyzed in the static setting, where the stream is assumed to be fixed in advance, and only then the randomnes...
[ לנוסח המלא ]

In this talk i'll be covering several works in the topic of 3D deep learning on pointclouds for scene understanding tasks. First, I'll describe VoteNet (ICCV 2019): a method for object detection from 3D pointclouds input, inspired by the classical generalized Hough voting technique. I'll then explain how we integrated image information into the voting scheme to further boost 3D detection (ImVoteNet, CVPR 2020). In the second part of my talk i'll describe a recent study about transfer-learnin...
[ לנוסח המלא ]

Mutation score is widely accepted to be a reliable measurement for the effectiveness of software tests. Recent studies, however, show that mutation analysis is extremely costly and hard to use in practice. We present a novel direct prediction model of mutation score using neural networks. Relying solely on static code features that do not require compilation or execution of the tests, we predict mutation score with an accuracy better than a quintile. When we include statement coverage as a featu...
[ לנוסח המלא ]

הנכם מוזמנים ליום עיון 2020 See Beyond Reality - שיתוף פעולה של חברת רפאל והפקולטה למדעי המחשב בטכניון אשר יתקיים ביום רביעי, 2 בדצמבר, 2020, החל מהשעה 09:00, בפגישה מקוונת. יום העיון יעסוק במחקרים מתחום Image Processing, Computer Vision, Computer Graphics פרטים נוספים בכרזה ה...
[ לנוסח המלא ]

We are living in an era where the digital world is becoming an inevitable part of our professional and daily lives. Digital creation tools are essential for many professions including design, entertainment, gaming etc. In our daily lives, we all take many pictures or capture many videos each day to record and share our memories. There is a stronger demand to transform such digital workflows into life-like experiences. My research focuses on enabling such a transformation by developing computatio...
[ לנוסח המלא ]

Motivated by applications in machine learning, such as subset selection and data summarization, we consider the problem of maximizing a monotone submodular function subject to mixed packing and covering constraints. We present a tight approximation algorithm that for any constant $\eps >0$ achieves a guarantee of $1-\nicefrac[ ]{1}{\e}-\eps$ while violating only the covering constraints by a multiplicative factor of $1-\eps$. Our algorithm is based on a novel enumeration method, which unli...
[ לנוסח המלא ]

In the "20 questions" game Alice picks a probability distribution over the set {1,...,n}, and draws a secret element x from it. Bob's goal is to construct a strategy to reveal x using only "Yes or No" questions, while minimizing the expected number of questions asked. Bob's success is guaranteed if he uses Huffman code, but then he might ask any possible question. In the paper "Twenty (simple) questions", Dagan et al. showed that the minimal number of questions which allows to construct optim...
[ לנוסח המלא ]

Genome sequencing data is rising exponentially at a rate (125%/year) that is far higher than our current transistor performance scaling (currently 3%/year). New medical and comparative genomics applications have emerged that ensure that the demand for more sequencing will continue to rise, which threatens to overwhelm our current compute capacities. Domain-specific acceleration (DSA), i.e. using specialized hardware for accelerating a narrow domain of algorithms, will enable us to tap the vast p...
[ לנוסח המלא ]

Image sensors capable of detecting individual photons are typically used in specialized low-light imaging applications. In this talk I will make a case for using time-resolved single-photon sensors as general-purpose image sensors, not limited to photon-starved scenarios. This may seem counterintuitive. Why would one want to use a single-photon sensor for situations other than imaging in low light? What additional information can a single-photon sensor provide that a conventional camera sensor c...
[ לנוסח המלא ]

Kirigami, the traditional Japanese art of paper cutting and folding generalizes origami and has initiated new research in material science as well as graphics. Jere we use its capabilities to perform geometric modeling with corrugated surface representations possessing an isometric unfolding into a planar domain after appropriate cuts are made. We initialize our box-based kirigami structures from orthogonal networks of curves, compute a first approximation of their unfolding via mappings ...
[ לנוסח המלא ]

הנכם מוזמנים לתחרות CTF- Capture The Flag הפקולטית הראשונה בנושאים שונים בתכנות, אבטחת מידע, אלגוריתמיקה, בינה מלאכותית ועוד. התחרות תיערך בקבוצות שיתמודדו עם אתגרים שיוצבו מטעם חברות גדולות בתעשייה, חברי סגל, סטודנטים ובוגרים בימים שישי עד ראשון, ...
[ לנוסח המלא ]

The notion of list-decoding, introduced by Elias in the 1950s, is a natural generalization of unique decoding, where the decoder is allowed to output a list of codewords that contains the transmitted codeword instead of a single codeword. Besides being a fundamental concept in coding theory, list decoding has found diverse applications in theoretical computer science. Error-correcting codes of rate R that are list-decodable up to the relative radius 1−R−ε are said to achieve the ...
[ לנוסח המלא ]

State sharding is a common solution to the scalability problem in blockchain systems, allowing nodes to hold a partial view of the system state. With sharding, the processing of a transaction might not be completed locally within a node and potentially requires involvement of multiple shards. Such cross-shards transactions have a negative impact on system performance and are frequent with traditional state partition solutions which are often based on a simple mapping of data into shards. By loca...
[ לנוסח המלא ]

In this talk we explore two contributions for shape analysis. In a first case, we consider surfaces and how local analysis of the angular oscillations and polynomial radial behavior around surface points leads to accurate normal estimation and new integral invariants. A direct application of these integral invariants is geometric detail exaggeration. This however assumes that the shape can be represented as a height function over some parameterization plane in a neighborhood of fixe...
[ לנוסח המלא ]

RNNs are a class of Machine learning models used to solve tasks with sequential structure. Since most neural circuits are recurrent, RNNs are often deployed to explain neural activity during computational tasks with a temporal dimension. Unlike their natural counterparts, an RNN’s entire computation is accessible. Yet, most works treat them as a black box and analyze their generated activity directly. A recent line of work uses RNNs as a hypothesis generation tool; The activity of a traine...
[ לנוסח המלא ]

In this paper we introduce randomized branching as a tool for parameterized approximation and develop the mathematical machinery for its analysis. Our algorithms improve the best known running times of parameterized approximation algorithms for Vertex Cover and 3-Hitting Set for a wide range of approximation ratios. One notable example is a simple parameterized random 1.5-approximation algorithm for Vertex Cover, whose running time of O*(1.01657k) substantially improves the best known running ...
[ לנוסח המלא ]

Monitoring network traffic is an important building block for various management and security systems. In typical settings, the number of active flows in a network node is much larger than the number of available monitoring resources and there is no practical way to maintain "per-flow" state at the node. This situation gave rise to the recent interest in streaming algorithms where complex data structures are used to perform monitoring tasks like identifying the top-$k$ flows, Heavy Hitter (HH) f...
[ לנוסח המלא ]

A major bottleneck in the evaluation of music generation is that music appreciation is a highly subjective matter. When considering an average appreciation as an evaluation metric, user studies can be helpful. The challenge of generating personalized content, however, has been examined only rarely in the literature. In this paper, we address generation of personalized music and propose a novel pipeline for music generation that learns and optimizes user-specific musical taste. We focus on the ta...
[ לנוסח המלא ]

A channel in the Lightning Network allows two parties to secure bitcoin payments and escrow holdings between them. Designed to increase transaction immediacy and reduce blockchain congestion, this has the potential to solve many issues associated with Bitcoin. In this talk, we study the economics of the Lightning Network. We present conditions under which two parties optimally establish a channel and give explicit formulas for channels’ costs. Using these, we derive implications for t...
[ לנוסח המלא ]

A common observation in data-driven applications is that high dimensional data has a low intrinsic dimension, at least locally. Thus, when one wishes to work with data that is not governed by a clear set of equations, but still wishes to perform statistical or other scientific analysis, an optional model is the assumption of an underlying manifold from which the data was sampled. This manifold, however, is not given explicitly but we can obtain samples of it (i.e., the individual data points). I...
[ לנוסח המלא ]

The potential benefits of taking surface curvature into account in engineering design have not yet been fully utilized. One of the problems is that differential geometry is often not offered in the undergraduate engineering program. In this presentation, I focus on a couple of research topics that demonstrate how curvatures play important roles in engineering design. The first topic is the effect of curvature on the energy absorption characteristics of cylindrical corrugated tubes under compress...
[ לנוסח המלא ]

Quotient spaces are a natural mathematical tool to describe a variety of algorithmic problems in computer vision and related fields, where different objects are to be compared while their natural symmetries are to be ignored. One popular approach for working in quotient spaces is cancelling out symmetries by finding a canonical alignment of the objects at hand. This approach typically involves solving a challenging optimization problem over the symmetry group. We will review the tractab...
[ לנוסח המלא ]

הנכם מוזמנים לסדנת "חשיפה למשרות בתעשייה - עושים סדר ב-Buzz Words" אשר מטרתה לסווג את המשרות בתעשייה, לפרש את מושגיה ומוצריה ולהמחיש יום עבודה טיפוסי בתפקידים השונים שהיא מציעה. הסדנה תתקיים באופן מקוון, בשני מפגשים: המפגש הראשון יתקיים ביום חמישי, 5 בנו�...
[ לנוסח המלא ]

In a polynomial protocol a prover sends messages that are polynomials, and the verifier is allowed to check polynomial identities between these polynomials. The prover complexity is measured as the sum of degrees of the polynomials sent. The motivation for the definition is to capture prover complexity in zero knowledge proofs systems based on polynomial commitment schemes. We will present and illustrate this notion; and present an open question on improved protocols for “range proofs...
[ לנוסח המלא ]

הנכם מוזמנים להרצאה ושיחה עם איתי רוזנבלט, ראש קבוצה ומוביל טכני ב-scaleIO ומייסד שותף וסמנכ"ל מו"פ בסטארט-אפ חדש, אשר יעסקו בנושא הראיון הטכני ובמרכיבי ההצלחה והרושם הראשוני שהוא מעניק: • מה המראיין מחפש בכל חלק מהראיון • מה ההזדמנות שלך כמועמד • איך להת...
[ לנוסח המלא ]

The complexity of Iterated Matrix Multiplication is a central theme in Computational Complexity theory, as the problem is closely related to the problem of separating various complexity classes within P. In this talk, we will talk about the complexity of computing an entry in the product of d many n×n generic matrices (denoted IMMn,d) by depth four circuits of bounded individual degree. We show that for all r ≤ na for a fixed constant a, any “syntactic” depth four ci...
[ לנוסח המלא ]

יום הכוון בפקולטה למדעי המחשב תשפ"א, 2020-21, לסטודנטים חדשים יתקיים באופן מקוון באמצעות שידור זום, ביום שלישי, 20 באוקטובר 2020, בשעה 10:30, ויכלול הרצאות מפי דיקן הפקולטה, סגן דיקן ללימודי תואר ראשון, נציגי ועד הסטודנטים ועוד, ולאחריהן יתקיימו מפגשי שאלות ותשו�...
[ לנוסח המלא ]

כולם מוזמנים לטקס הענקת תארים לבוגרי מחזור צ"א שיתקיים בשידור ישיר ביוטיוב היום, יום חמישי, 15 באוקטובר 2020, בשעה 18:30. מצפים לכם בהתרגשות!...
[ לנוסח המלא ]

The planted clique problem plays an important role in the field of average-case complexity. The problem consists of detecting graphs that contain a hidden clique of size $k$ that was planted in graphs from the Erdos-Renyi model $G_{n,0.5}$. The planted clique conjecture suggests that there is no polynomial-time algorithm that can solve the planted clique problem when $2\log_2(n) \ll k \ll \sqrt{n}$. This conjecture has served as the underlying hardness assumption in several different works i...
[ לנוסח המלא ]

Reliability and time-efficiency are two key elements to consider in network design. Commonly, each is measured per service - availability probability of a specific service, the latency of a specific service, and overall - system average reliability and system average latency, considering the demand for every service. Intuitively, minimizing latency requires minimizing the number of network elements a service makes use of. In a non-redundant environment, this would also guarantee the maximal reli...
[ לנוסח המלא ]

A statistical zero-knowledge proof for a problem Pi enables a computationally unbounded prover to convince a polynomial-time verifier that x belongs to Pi without revealing any additional information about x to the verifier, in a strong information-theoretic sense. Suppose, however, that the prover wishes to convince the verifier that k separate inputs x_1,...,x_k all belong to Pi (without revealing anything else). A naive way of doing so is to simply run the SZK protocol separately for each ...
[ לנוסח המלא ]

In blockchain applications, transaction latency is a key factor in the quality of service (QoS). Each application that uses the blockchain network has the motivation to reduce the latency for its own transactions. The block proposer selects which pending transaction to include in the proposed block and would like to prioritize its own application transactions over those of other applications. To maintain fairness, a block proposer can be asked to select the transactions randomly among pending tr...
[ לנוסח המלא ]

The fields of signal and image processing have been deeply influenced by the introduction of deep neural networks. These are successfully deployed in a wide range of real-world applications, obtaining state of the art results and surpassing well-known and well-established classical methods. Despite their impressive success, the architectures used in many of these neural networks come with no clear justification. As such, these are usually treated as ``black box'' machines that lack any kind of i...
[ לנוסח המלא ]

A functional PIR array code is a coding scheme which encodes some s information bits into a tXm array such that every linear combination of the s information bits has k mutually disjoint recovering sets. Every recovering set consists of some of the array's columns while it is allowed to read at most l encoded bits from every column in order to receive the requested linear combination of the information bits. Functional batch array codes impose a stronger property where every multiset request of ...
[ לנוסח המלא ]

We tackle the task of Term Set Expansion (TSE): given a small seed set of example terms from a semantic class, finding more members of that class. The task is of great practical utility, and also of theoretical utility as it requires generalization from few examples. Previous approaches to the TSE task can be characterized as either distributional or pattern-based. We harness the power of neural masked language models (MLM) and propose a novel TSE algorithm, which combines the pattern-based and ...
[ לנוסח המלא ]

Shay Moran, Technion CS graduate, will give a lecture on his work in the theory of learning....
[ לנוסח המלא ]

A new family of codes, called clustering-correcting codes, is presented in this paper. This family of codes is motivated by the special structure of the 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, an important task in the decodin...
[ לנוסח המלא ]

Machine learning has made considerable progress over the past decade, matching and even surpassing human performance on a varied set of narrow computational tasks. This progress has been enabled by the widespread availability of large datasets, as well as by improved algorithms and models. Distribution, implemented either through single-node concurrency or through multi-node parallelism, has been the third key ingredient to these advances. The goal of this talk is to provide an overvi...
[ לנוסח המלא ]

טקס חלוקת תעודות לבוגרי תואר שלישי בטכניון יתקיים באופן מקוון הערב, יום רביעי, 16 בספטמבר 2020, וישודר בפייסבוק. השנה יוענקו תעודות למספר מכובד של בוגרים - 231 בוגרי הטכניו�...
[ לנוסח המלא ]

מרכז המחקר לאבטחת סייבר ע"ש הירושי פוג'יווארה יקיים את בית-ספר קיץ השמיני על סייבר ואבטחת מחשבים בנושא: "Privacy in Challenging Times" הכנס יתקיים בימים ב'-ה', 10-7 בספטמבר, 2020, בשידור זום הטכניון. דוברים: Keynote Speakers: Ro...
[ לנוסח המלא ]

A proof of work (PoW) blockchain protocol distributes rewards to its participants, called miners, according to their share of the total computational power. Sufficiently large miners can perform selfish mining - deviate from the protocol to gain more than their fair share. Such systems are thus secure if all miners are smaller than a threshold size so their best response is following the protocol. To find the threshold, one has to identify the optimal strategy for miners of different sizes, i...
[ לנוסח המלא ]

הנכם מוזמנים להרצאה ושיחה עם דר' יונתן יניב, בוגר הפקולטה וראש קבוצת המחקר בחברת YOTPO, אשר יעסקו בשאלות: האם כדאי לעבוד במהלך התואר? כיצד להשתלב בתעשייה, בייחוד עכשיו בתקופת הקורונה? כיצד מוצאים את המשרה המתאימה לכם? כיצד בונים מסלול קריירה? עבודה בחברה גדולה...
[ לנוסח המלא ]

The Shapley value is a conventional and well-studied function for determining the contribution of a player to the coalition in a cooperative game. Among its applications in a plethora of domains, it has recently been proposed to use the Shapley value for quantifying the contribution of a tuple to the result of a database query. In particular, we have a thorough understanding of the tractability frontier for the class of Conjunctive Queries (CQs) and aggregate functions over CQs. It has also b...
[ לנוסח המלא ]

הנכם מוזמנים להשתתף בסדנה וירטואלית ב נושא מערכות לומדות ואבטחת חומרה שתתקיים בימים רביעי וחמישי, 26-27 באוגוסט, 2020, בשידור בזום....
[ לנוסח המלא ]

Scene Text Recognition (STR), the task of recognizing text against complex image backgrounds, is an active area of research. Current state-of-the-art (SOTA) methods still struggle to recognize text written in arbitrary shapes. In this paper, we introduce a novel architecture for STR, named Selective Context ATtentional Text Recognizer (SCATTER). SCATTER utilizes a stacked block architecture with intermediate supervision during training, that paves the way to successfully train a deep BiLSTM enco...
[ לנוסח המלא ]

Deduplication decreases the physical occupancy of files in a storage volume by removing duplicate copies of data chunks,but creates data-sharing dependencies that complicate standard storage management tasks. Specifically, data migration plans must consider the dependencies between files that are remapped to new volumes and files that are not. Thus far, only greedy approaches have been suggested for constructing such plans, and it is unclear how they compare to one another and how much they can ...
[ לנוסח המלא ]

We inspect the fine-grained complexity of answering queries over relational databases. With the ideal guarantees, linear time is required before the first answer to read the input and determine its existence, and then we need to print the answers one by one. Consequently, we wish to identify the queries that can be solved with linear preprocessing time and constant or logarithmic delay between answers. A known dichotomy classifies CQs into those that admit such enumeration and those that do no...
[ לנוסח המלא ]

Optical character recognition (OCR) systems performance have improved significantly in the deep learning era. This is especially true for handwritten text recognition (HTR ), where each author has a unique style, unlike printed text, where the variation is smaller by design. That said, deep learning based HTR is limited, as in every other task, by the number of training examples. Gathering data is a challenging and costly task, and even more so, the labeling task that follows, of which we focus ...
[ לנוסח המלא ]

King and Saia were the first to break the quadratic word complexity bound for Byzantine Agreement in synchronous systems against an adaptive adversary, and Algorand broke this bound with near-optimal resilience (first in the synchronous model and then with eventual-synchrony). Yet the question of asynchronous sub-quadratic Byzantine Agreement remained open. To the best of our knowledge, we are the first to answer this question in the affirmative. A key component of our solution is a shared coin ...
[ לנוסח המלא ]

Although Magnetic Resonance Imaging (MRI) is a superb medical imaging modality, its clinical use is limited by its long acquisition time. The acquisition time can be shortened by sampling data with a sub-Nyquist rate; however, this requires suitable methods for solving the ill-posed inverse problem of accurate image reconstruction from highly subsampled data. In this seminar, four novel methods for solving the MRI reconstruction problem will be presented. These methods build upon th...
[ לנוסח המלא ]

We report a series of robust empirical observations, whereby deep Neural Networks learn the examples in both the training and test sets in a similar order. This phenomenon is observed in all the commonly used benchmarks we evaluated, including many image classification benchmarks, and one text classification benchmark. While this phenomenon is strongest for models of the same architecture, it also crosses architectural boundaries -- models of different architectures start by learning the same ex...
[ לנוסח המלא ]

In this talk, I will describe a couple of results from my PhD thesis, that studies distributed algorithms for connectivity and distance related graph problems. I will focus on a recent line of work studying distance computation in the Congested Clique model of distributed computing. We design fast algorithms for the fundamental all-pairs shortest paths (APSP) and multi-source shortest paths (MSSP) problems. In particular, we show constant-factor approximation algorithms for weighted APS...
[ לנוסח המלא ]

One of the fascinating properties of deep learning is the ability of the network to reveal the underlying factors characterizing elements in datasets of different types. Autoencoders represent an effective approach for computing these factors and also have been studied in the context of their ability to interpolate between data points by decoding mixed latent vectors. This interpolation often incorporates disrupting artifacts or produces unrealistic images during reconstruction. We argue that th...
[ לנוסח המלא ]

Finding exact solutions to many fundamental optimization and distance computation problems is known to be a hard task in the distributed CONGEST model of computation. A natural relaxation then is to look for approximate solutions. In this talk, I will discuss some of the research I did during my M.Sc. studies that concerns the trade-off between approximation ratio and round complexity for central optimization and distance computation problems....
[ לנוסח המלא ]

In the first part of this two-part talk, I will describe my work in adversarial machine learning (ML). Prior work has demonstrated that ML algorithms are vulnerable to evasion attacks at inference time by so-called adversarial examples. However, the implications of attacks in practice, under real-world constraints, remained largely unknown. To fill the gap, we propose evasion attacks that satisfy multiple objectives (e.g., smoothness and robustness against changes in imaging conditions), and sho...
[ לנוסח המלא ]

In the directed minimum spanning tree problem (DMST, also called minimum weight arborescence), the network is given a root node r, and needs to construct a minimum-weight directed spanning tree, rooted at r and oriented outwards. In this paper we present the first sub-quadratic DMST algorithms in the distributed CONGEST network model, where the messages exchanged between the network nodes are bounded in size. We consider three versions: a model where the communication links are bidirectional but...
[ לנוסח המלא ]

הפקולטה למדעי המחשב מזמינה אתכם למפגש היכרות עם Flutter - טכנולוגיית אפליקציות חדשה של גוגל - באמצעות ZOOM, ביום רביעי, 10 ביוני, 19:00.בהשתתפות: ברק מנחם, מנהל קהילת Flutter Israel Developers תמיר חיים רביע, מומחה Flutter, React Native בתוכנית: טכנולוגיות Cross-platform מהי Flutter ומד...
[ לנוסח המלא ]

יום העיון בסייבר ואבטחת מידע 2020 באירגונם של אלי ביהם, שרה ביתן ומרכז המחקר לאבטחת סייבר ע"ש הירושי פוג'יאוורה, יתקיים במרחב הסייבר (באמצעות זום - קישור יישלח בסמוך לאירוע לאחר הרשמה), ביום רביעי, 10 ביוני 2020. ...
[ לנוסח המלא ]

Automated tools and algorithms for synthesizing geometry have been ubiquitous in computer-aided design systems, since their very inception. Beyond the simple example of sketching tools for designers, algorithms for synthesizing geometry take many forms, such as design-by-constraints, optimization, and toolpath generation. In this thesis, we will present three problems in the field on geometry processing, and examine them both within the context of geometry synthesis, and in their broader applica...
[ לנוסח המלא ]

הפקולטה למדעי המחשב מזמינה תלמידי תארים מתקדמים למפגש בנושא מחקר בתעשיית ההיטק: האם יש מחקר בתעשייה בישראל ומהם סוגי המשרות ומסלולי קריירה הקיימים, אשר יידון בפאנל עם: ליאן לוין איתן, מנהלת מחקר בכירה, אמאזון טל מזרחי, אדריכל וחוקר ראש...
[ לנוסח המלא ]

Computer system performance has improved due to creatively using more transistors (Moore’s Law) in parallel via bit-, instruction-, thread-, and data-level parallelism. With the slowing of technology scaling, a way to further improve computer system performance under energy constraints is to employ hardware accelerators. Each accelerator is a hardware component that executes a targeted computation class faster and usually with (much) less energy. Already today, many chips in mobile, edge and c...
[ לנוסח המלא ]

Licklider and Taylor (1968) envisioned computational machinery that could enable better communication between humans than face-to-face. In the last fifty years, we have used computing to develop various means of communication, such as mails, messaging, video conversation, or virtual reality. These are, however, a proxy of face-to-face communication that aims at encoding words, expressions, emotions, and body language at source and try to decode them reliably at the destination. The true potentia...
[ לנוסח המלא ]

Optimization in machine learning, both theoretical and applied, is presently dominated by first-order gradient methods such as stochastic gradient descent. Second-order optimization methods---that involve second-order derivatives and/or second-order statistics of the data---have become far less prevalent despite strong theoretical properties, due to their impractical computation, memory and communication costs. I will present some recent theoretical, algorithmic and infrastructu...
[ לנוסח המלא ]

The idea of light field technology is relatively old, however, apart from various attempts of construction of imaging devices which can densely capture scene radiance, only recent availability of devices from companies like Lytro and Raytrix started a significant increase of interest in the light field technology. The core idea of the light field is in the dense sampling of viewpoints where the dense collection of views distinguishes is from the sparse sampling in multiview stereo. Having access...
[ לנוסח המלא ]

In this work, we present a method for node embedding in temporal graphs. We propose an algorithm that learns the evolution of a temporal graph's nodes and edges over time and incorporates this dynamics in a temporal node embedding framework for different graph prediction tasks. We present a joint loss function that creates a temporal embedding of a node by learning to combine its historical temporal embeddings, such that it optimizes per given task (e.g., link prediction). The algorithm is initi...
[ לנוסח המלא ]

Two-player turn-based zero-sum stochastic games is an interesting family of infinite duration games played on a finite set of states. These games may be seen as an extension of Markov Decisions Processes (MDPs) to a two-player setting. Such games have applications in various areas ranging from Artificial Intelligence, Operations Research and Automatic Verification. From the theoretical point of view they are interesting because they are known to be in NP intersection co-NP, yet no polynomial...
[ לנוסח המלא ]

Hypernetworks, also known as dynamic networks, are neural networks in which the weights of at least some of the layers vary dynamically based on the input. Such networks have composite architectures in which one network predicts the weights of another network. I will briefly describe the early days of dynamic layers and present recent results from diverse domains: 3D reconstruction from a single image, image retouching, electrical circuit design, decoding block codes, graph hypernetworks for b...
[ לנוסח המלא ]

האירוע יתקיים בשידור חי באמצעות Z00M  - מס' פגישה: 518018934 היום הפתוח מזמין בוגרי תואר ראשון מצטיינים מכל האוניברסיטאות להירשם ללימודי תארים מתקדמים בפקולטה למדעי המחשב בסמטר חורף תשפ"א. ...
[ לנוסח המלא ]

Voice assistants are designed to make our lives easier. By responding to spoken commands and questions, they provide a natural way of interaction with computers and smartphones. The problem starts when these voice assistants run by default even when the device is locked, which requires vendors to balance comfort with security. In this talk, I will take the audience on an amusing journey of discovering more than twenty new security vulnerabilities in popular voice assistants, and the fascinating ...
[ לנוסח המלא ]

Deep Learning has always been divided into two phases: Training and Inference. Deep networks are mostly used with large data-sets, both under supervised (Classification, Regression etc.) or unsupervised (Autoencoders, GANs) regimes. Such networks are only applicable to the type of data they were trained for, and do not exploit the internal statistics of a single datum. We introduce Deep Internal Learning; We train a signal-specific network, we do it at test-time and on the test-inpu...
[ לנוסח המלא ]

Recently, there is an increase in delegating computations to powerful remote servers. This appearance raises questions of computational integrity and efficient verification. For instance, the server can be unreliable, therefore it may output erroneous results. In other contexts, where preserving information privacy is crucial, as in medical and forensic data, other issues might arise. The veils of secrecy, that are designed to keep the data hidden from the public, may be abused to cover up lies ...
[ לנוסח המלא ]

Contemporary presentation software provides sufficient tools for simple presentation scenarios. However, such tools have not changed for about two decades and are unsuitable for many complex modern needs. Content is typically limited to text and images only and cannot be expressed well in terms of 2D slides and bullets quantization. Such limitations are ill-suited for many modern needs where linear sequences of fixed-sized 2D slides consisting of text and images is not nearly sufficient. ...
[ לנוסח המלא ]

Contemporary presentation software provides sufficient tools for simple presentation scenarios. However, such tools have not changed for about two decades and are unsuitable for many complex modern needs. Content is typically limited to text and images only and cannot be expressed well in terms of 2D slides and bullets quantization. Such limitations are ill-suited for many modern needs where linear sequences of fixed-sized 2D slides consisting of text and images is not nearly sufficient. In t...
[ לנוסח המלא ]

Online Portfolio Selection, aiming to optimize the allocation of wealth across a set of assets, is a fundamental research problem in computational finance and machine learning. Despite the theoretical challenges, the implementation of a real-world trading system is extremely challenging. This issue has been extensively studied across several research communities, including finance, statistics, coding and information theory and machine learning. This long-standing problem, however, still poses ma...
[ לנוסח המלא ]

Applications such as distributed learning and edge computing strive to maximize the number of service requests (e.g., for data access) that can be concurrently executed by the system. Redundancy, in the form of simple replication and erasure coding, has emerged as an efficient and robust way to enable simultaneous access of different data objects by many users competing for the system’s resources. Here, replication and coding of data affect the rates at which users can be simultaneously served...
[ לנוסח המלא ]

Bluetooth is a widely deployed standard for wireless communications between mobile devices. It uses authenticated Elliptic Curve Diffie-Hellman for its key exchange. In this paper we show that the authentication provided by the Bluetooth pairing protocols is insufficient and does not provide the promised MitM protection. We present a new attack that modifies the y-coordinates of the public keys (while preserving the x-coordinates). The attack compromises the encryption keys of all of the current...
[ לנוסח המלא ]

In this talk, we will discuss recent advances in multitask regression and their application to flood forecasting. We will begin with a brief overview of Google’s flood forecasting initiative in India. The project involves time series prediction in different geographical locations based on a limited number of heavy tailed samples. To address these challenges, we will consider non-convex robust multitask models based on elliptical distributions. Next, we will present a spectral algorithm for lea...
[ לנוסח המלא ]

Magnetic Resonance Imaging (MRI) has long been considered to be among "the gold standards" of diagnostic medical imaging. The long acquisition times, however, render MRI prone to motion artifacts, let alone their adverse contribution to the relative high costs of MRI examination. Over the last few decades, multiple studies have focused on the development of both physical and post-processing methods for accelerated acquisition of MRI scans. These two approaches, however, have so far been address...
[ לנוסח המלא ]

The manual detection, examination and repair of computer bugs are all notoriously difficult tasks that programmers face daily. Automated program repair receives a program with a bug, and outputs a set of changes to the program that would make it bug-free. We focus on formal program repair, which means that programs are repaired with respect to a formal specification. Formal program repair, in contrast to test-based repair, guarantees that programs meet the specification for all inputs, and not j...
[ לנוסח המלא ]

Medical ultrasound is a widespread imaging modality owing its popularity to cost efficiency, portability, speed, and lack of harmful ionizing radiation. In this work, we demonstrate that replacing the traditional ultrasound processing pipeline with a data-driven, learnable counterpart leads to significant improvement in image quality. Moreover, we demonstrate that greater improvement can be achieved through a learning-based design of the transmitted beam patterns simultaneously with learning an ...
[ לנוסח המלא ]

The recent and on-going Deep-Learning (DL) revolution, introduces a paradigm shift in almost all disciplines of signal processing. Traditionally, Computer Vision (CV) and Image Processing (IP) methods were based on hand-crafted feature extraction from the initial optical image, and then some hand-crafted classifier/filter was defined to achieve the desired result IP/CV. The machine learning approach seeks to learn the 'classifier' stage using data (either labeled or unlabeled), i.e. the decision...
[ לנוסח המלא ]

We introduce a novel neural network-based partial differential equations solver for forward and inverse problems. The solver is grid free, mesh free and shape free, and the solution is approximated by a neural network. We employ an unsupervised approach such that the input to the network is a points set in an arbitrary domain, and the output is the set of the corresponding function values. The network is trained to minimize deviations of the learned function from the PDE solution ...
[ לנוסח המלא ]

As data analytics becomes more crucial to digital systems, so grows the importance of characterizing the database queries that admit a more efficient evaluation. We consider the tractability yardstick of answer enumeration with a polylogarithmic delay after a linear-time preprocessing phase. Such an evaluation is obtained by constructing, in the preprocessing phase, a data structure that supports polylogarithmic-delay enumeration. In this paper, we seek a structure that supports the more demandi...
[ לנוסח המלא ]

We are interested in the following question: How could one create a digital personal assistant, capable of acquiring and applying new skills through spoken interaction? The most rudimentary requirement of such skill acquisition would be Turing complete programming. Other such acquisitions might be the deduction of routines from user actions and their automation thereof. Modern digital personal assistants do not possess the ability to learn through Voice User Interface (VUI), which is their pr...
[ לנוסח המלא ]

The recent push towards ultra-high (magnetic) fields in Magnetic Resonance Imaging (MRI) is expected to change the face of biomedical imaging. Recent studies employing state of the art 7 T MRI scanners achieved submillimeter resolution in 3D anatomical imaging of the human brain. However, the higher field also comes with new challenges, especially since we want to acquire the images with high spatial or temporal resolution to study the human brain function. These challenges include complex RF fi...
[ לנוסח המלא ]

We prove that for all constants a, NQP = NTIME[n^polylog(n)] cannot be (1/2 + 2^(-log^a n) )-approximated by 2^(log^a n)-size ACC^0 circuits. Previously, it was even open whether E^NP can be (1/2+1/sqrt(n))-approximated by AC^0[2] circuits. As a straightforward application, we obtain an infinitely often non-deterministic pseudorandom generator for poly-size ACC^0 circuits with seed length 2^{log^eps n}, for all eps > 0. More generally, we establish a connection showing that, for a ty...
[ לנוסח המלא ]

Reconstructing observed images from fMRI brain recordings is challenging. Unfortunately, acquiring sufficient "labeled" pairs of {Image, fMRI} (i.e., images with their corresponding fMRI responses) to span the huge space of natural images is prohibitive for many reasons. We present a novel approach which, in addition to the scarce labeled data (training pairs), allows to train fMRI-to-image reconstruction networks also on "unlabeled" data (i.e., images without fMRI recording, and fMRI recording ...
[ לנוסח המלא ]

Suppose that in a given planar circular region, there are some smart mobile evaders and we would like to find them using swarms of sweeping agents. A smart evader is a target that detects and responds to the motions of searchers by performing evasive maneuvers, to avoid interception. We assume various search configurations for the sweeping swarm of agents, and present guaranteed search techniques for single agent and multi agent swarms. These search procedures enable both confineme...
[ לנוסח המלא ]

In this talk I will review some major advances in the theory of distributed computing in the past decade and discuss key research challenges for the 2020s, with the main focus on distributed graph algorithms. I will present promising new approaches for doing research in the field, and I will also highlight examples of seemingly elementary questions that are still beyond the reach of our current techniques. ...
[ לנוסח המלא ]

מעבדות הפקולטה למדעי המחשב: המעבדה לפיתוח תוכנה ומערכות (ICST), המעבדה לסייבר ואבטחת מידע (CYBER) ותקשורת מחשבים (LCCN), והמעבדה לעיבוד גיאומטרי של תמונות (GIP), מזמינות אתכם לבקר ביריד האביב של פרוייקטים בתחומי IoT, תוכנה, אפל...
[ לנוסח המלא ]

Least-mean squares (LMS) solvers such as Linear / Ridge / Lasso-Regression, SVD and Elastic-Net not only solve fundamental machine learning problems, but are also the building blo cks in a variety of other methods, such as decision trees and matrix factorizations. We suggest an algorithm that gets a finite set of n d-dimensional real vectors and returns a weighted subset of d + 1 vectors whose sum is exactly the same. The proof in Caratheodory’s Theorem (1907) computes such a sub...
[ לנוסח המלא ]

We propose a method to automatically match two shape collections with a similar shape space structure, e.g. two characters in similar poses, and compute the inter-maps between the collections. Given the intra-maps in each collection, we extract the corresponding shape difference operators, and use them to construct an embedding of the shape space of each collection. We then align the two shape spaces, and use the knowledge gained from the alignment to compute the inter-maps. Unlik...
[ לנוסח המלא ]

The task of synthetic data generation can be, roughly, stated as follows: A learner gets to observe examples, sampled from some unknown distributions, and needs to output "synthetic" examples that "look similar". For example, think of an algorithm that receives examples of music tunes from some genre (e.g. jazz, classical, etc.) and in turn generates new synthetic pieces of music from that genre. This task has won considerable attention lately, due to the introduction of the algorithmic framewor...
[ לנוסח המלא ]

We address the problem of binary code search in stripped executables (with no debug information). The main challenge is establishing binary code similarity even when the binary code has been compiled using different compilers, optimization levels and targeting diverse architectures. Moreover, the source code being compiled might be from another version of the software package or another implementation altogether. Overcoming these challenges, while avoiding false-positives, is invaluable to guidi...
[ לנוסח המלא ]

The study of permutations is motivated by their application in coding for flash memories, and their relevance in different applications of networking technologies and various channels. We study the multidimensional constrained systems of $\mathbb{Z}^d$-permutations with restricted movement. During the talk, we will show a correspondence between these restricted permutations and perfect matchings. We will use the theory of perfect matchings to investigate several two-dimensional cases, for which ...
[ לנוסח המלא ]

Modern machine learning algorithms have achieved remarkable performance in a myriad of applications, and are increasingly used to make impactful decisions in the hiring process, criminal sentencing, healthcare diagnostics and even to make new scientific discoveries. The use of data-driven algorithms in high-stakes applications is exciting yet alarming: these methods are extremely complex, often brittle, notoriously hard to analyze and interpret. Naturally, concerns have raised about the reliabil...
[ לנוסח המלא ]

Inmodern applications of graph algorithms, where the graphs of interest arelarge and dynamic, it is unrealistic to assume that an input representationcontains the full information of a graph being studied. For example, consider asocial network, where a vertex corresponds to a user and an edge corresponds toa friendship relation. It is reasonable to assume that users do not alwaysupdate new friendships on the social network, and that sometimes theydo not fully disclose their friendship relations ...
[ לנוסח המלא ]

We are living in the era of massive data. Some estimate that more than 90% of the worlds data was generated in recent years. This large amount of data comes with technological challenges such as: distributed storage systems, synchronization issues, density constraints, power constraints and more. The talk will be partitioned into two, independent parts in which some solutions to those challenges will be discussed. Specifically, we will touch upon dynamical distributed storage systems, and Polya...
[ לנוסח המלא ]

We study the online problems related to controlling a linear time-invariant systems with noisy dynamics where either: (1) onlinevadversarially chosen quadratic losses and known dynamics, or (2) unknown dynamics but known quadratic losses. We present an efficient online learning algorithms for both setting with vanishing regret. Based on joint works with Alon Cohen, Avinatan Hassidim, Tomer Koren, Nevena Lazic, and Kunal Talwar Bio: Prof. Yishay Mansour got...
[ לנוסח המלא ]

Formal verification of infinite-state systems, and distributed systems in particular, is a long standing research goal. I will describe a series of works that develop a methodology for verifying distributed algorithms and systems using decidable logics, employing decomposition, abstraction, and user interaction. This methodology is implemented in an open-source tool, and has resulted in the first mechanized proofs of several important distributed protocols. I will also describe a novel approach ...
[ לנוסח המלא ]

This talk presents new computational and statistical barriers in machine learning, along with the algorithmic developments that they inspire. The computational barriers arise in nonconvex optimization: we prove lower bounds on the (oracle) complexity of finding stationary points using (stochastic) gradient methods, showing that gradient descent is unimprovable for a natural class of problems. We bypass this barrier by designing an algorithm that outperforms gradient descent for a large subclass...
[ לנוסח המלא ]

We propose a method for tracing the flowlines of a discrete tangent vector field on a triangle mesh. Our method makes use of functional vector fields, namely modeling of vector fields as operators acting on functions. In particular, we use the recently proposed adjoint representations of directional derivatives. Our method is characterized by making use of mostly global solutions, as opposed to iterative local algorithms that trace the vector triangle-by-triangle. We compare our approach to ana...
[ לנוסח המלא ]

We propose a method for tracing the flowlines of a discrete tangent vector field on a triangle mesh. Our method makes use of functional vector fields, namely modeling of vector fields as operators acting on functions. In particular, we use the recently proposed adjoint representations of directional derivatives. Our method is characterized by making use of mostly global solutions, as opposed to iterative local algorithms that trace the vector triangle-by-triangle. We compare our approach to ana...
[ לנוסח המלא ]

Compilers are designed today to use fixedAbstract -cost models that are based on heuristics to make different code optimizations. However, these models are unable to capture the data dependency, the computation graph, or the organization of instructions. Therefore, with these models compilers achieve suboptimal performance. In his talk, Ameer will show how to use deep reinforcement learning to overcome such hard compiler challenges. This includes the NP-Hard compiler phase ordering challenge (Au...
[ לנוסח המלא ]

We develop differentially private mechanisms that achieve nearly instance-optimal losses, achieving lower loss than all appropriately unbiased mechanisms for any possible instance. We show that our mechanisms, with a modest increase in sample size (logarithmic or constant), are instance-optimal for a large family of functions. In contrast to existing mechanisms, which use the global or local sensitivity of the function being estimated, and so are necessarily instance suboptimal, the key to our c...
[ לנוסח המלא ]

One of the unique aspects of quantum mechanics is the no-cloning theorem. This theorem has direct applications in quantum crpytography: quantum money is very similar to the cash we use, but in addition, it is physically impossible to forge. In this talk, I will discuss quantum money, and some of its extensions (quantum coins, semi quantum money, tokens for digital signatures, quantum copy protection, and quantum lightning). I will also descrube a way to use quantum money in a system s...
[ לנוסח המלא ]

Dynamical systems are everywhere, from the flow of particles in the air to the evolution of people's personalities. Recently, the big data revolution has made increasing amounts of observations of such systems readily available. Unfortunately, in many practical scenarios, the governing equations of the dynamics are unknown. For instance, what would be the mathematical model for periodic questionnaire data that describes personalities? Despite the lack of existing models, scientists and engineers...
[ לנוסח המלא ]

This talk presents a novel end-to-end approach to program repair based on sequence-to-sequence learning. We devise, implement, and evaluate a system, called SequenceR, for fixing bugs based on sequence-to-sequence learning on source code. This approach uses the copy mechanism to overcome the unlimited vocabulary problem that occurs with big code. Our system is data-driven; we train it on 35,578 samples, carefully curated from commits to open-source repositories. We evaluate it on 4,711 indepen...
[ לנוסח המלא ]

We consider the problem of training and evaluating generative models with a Generative Adversarial Network (GAN). Although GANs can accurately model complex distributions, they are known to be difficult to train due to instabilities caused by a difficult minimax optimization problem. In this work, we view the problem of training GANs as finding a mixed strategy in a zero-sum game. This leads to a new training method that builds upon ideas from game theory and online learning. We provide theoret...
[ לנוסח המלא ]

Recent developments in Natural Language Processing (NLP) allow models to leverage large, unprecedented amounts of raw text, culminating in impressive performance gains in many of the field's long-standing challenges, such as machine translation, question answering, or information retrieval. In this talk, I will show that despite these advances, state-of-the-art NLP models often fail to capture crucial aspects of text understanding. Instead, they excel by finding spurious patterns in the data, w...
[ לנוסח המלא ]

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 enables flexible packet processing while optimizing for throughput and latency. In th...
[ לנוסח המלא ]