Deborah Cohen

Deborah Cohen

Deborah Cohen received the B.Sc. degree in electrical engineering (summa cum laude) in 2010 and the Ph.D. degree in electrical engineering (signal processing) in 2016 from the Technion - Israel Institute of Technology, Haifa, in 2010. Since 2010, she has been a Project Supervisor with the Signal and Image Processing Lab, the High Speed Digital Systems Lab, the Communications Lab and the Signal Acquisition, Modeling and Processing Lab (SAMPL), at the Electrical Engineering Department, Technion. In 2011, Ms. Cohen was awarded the Meyer Foundation Excellence prize. She received the Sandor Szego Award and the Vivian Konigsberg Award for Excellence in Teaching from 2012 to 2016, the David and Tova Freud and Ruth Freud-Brendel Memorial Scholarship in 2014 and the Muriel and David Jacknow Award for Excellence in Teaching in 2015. Since 2014, Ms. Cohen is an Azrieli Fellow. She is currently a research scientist in the Clair team in Google Israel. Her research interests include theoretical aspects of signal processing, compressed sensing, reinforcement learning and machine learning for dialogues.
Authored Publications
Google Publications
Other Publications
Sort By
  • Title
  • Title, descending
  • Year
  • Year, descending
    AI Increases Global Access to Reliable Flood Forecasts
    Asher Metzger
    Dana Weitzner
    Frederik Kratzert
    Guy Shalev
    Martin Gauch
    Sella Nevo
    Shlomo Shenzis
    Tadele Yednkachw Tekalign
    Vusumuzi Dube
    arXiv(2023)
    Preview abstract Floods are one of the most common natural disasters, with a disproportionate impact in developing countries that often lack dense streamflow gauge networks. Accurate and timely warnings are critical for mitigating flood risks, but hydrological simulation models typically must be calibrated to long data records in each watershed. Here we show that AI-based forecasting achieves reliability in predicting extreme riverine events in ungauged watersheds at up to a 5-day lead time that is similar to or better than the reliability of nowcasts (0-day lead time) from a current state of the art global modeling system (the Copernicus Emergency Management Service Global Flood Awareness System). Additionally, we achieve accuracies over 5-year return period events that are similar to or better than current accuracies over 1-year return period events. This means that AI can provide flood warnings earlier and over larger and more impactful events in ungauged basins. The model developed in this paper was incorporated into an operational early warning system that produces publicly available (free and open) forecasts in real time in over 80 countries. This work highlights a need for increasing the availability of hydrological data to continue to improve global access to reliable flood warnings. View details
    Preview abstract We study conversational domain exploration (CODEX), where the user’s goal is to enrich her knowledge of a given domain by conversing with an informative bot. Such conversations should be well grounded in high-quality domain knowledge as well as engaging and open-ended. A CODEX bot should be proactive and introduce relevant information even if not directly asked for by the user. The bot should also appropriately pivot the conversation to undiscovered regions of the domain. To address these dialogue characteristics, we introduce a novel approach termed dynamic composition that decouples candidate content generation from the flexible composition of bot responses. This allows the bot to control the source, correctness and quality of the offered content, while achieving flexibility via a dialogue manager that selects the most appropriate contents in a compositional manner. We implemented a CODEX bot based on dynamic composition and integrated it into the Google Assistant. As an example domain, the bot conversed about the NBA basketball league in a seamless experience, such that users were not aware whether they were conversing with the vanilla system or the one augmented with our CODEX bot. Results are positive and offer insights into what makes for a good conversation. To the best of our knowledge, this is the first real user experiment of open-ended dialogues as part of a commercial assistant system. View details
    Sparse imitation learning for text based games with combinatorial action spaces
    Chen Tessler
    Tom Zahavy
    Daniel J. Mankowitz
    Shie Mannor
    The Multidisciplinary Conference on Reinforcement Learning and Decision Making (RLDM)(2019)
    Preview abstract We propose a computationally efficient algorithm that combines compressed sensing with imitation learning to solve text-based games with combinatorial action spaces. Specifically, we introduce a new compressed sensing algorithm, named IK-OMP, which can be seen as an extension to the Orthogonal Matching Pursuit (OMP). We incorporate IK-OMP into a supervised imitation learning setting and show that the combined approach (Sparse Imitation Learning, Sparse-IL) solves the entire text-based game of Zork1 with an action space of approximately 10 million actions given both perfect and noisy demonstrations. View details
    Preview abstract Complex classifiers may exhibit ``embarassing'' failures in cases that would be easily classified and justified by a human. Avoiding such failures is obviously paramount, particularly in domains where we cannot accept such unexplained behavior. In this work we focus on one such setting, where a label is perfectly predictable if the input contains certain features, and otherwise, it is predictable by a linear classifier. We define a related hypothesis class and determine its sample complexity. We also give evidence that efficient algorithms cannot, unfortunately, enjoy this sample complexity. We then derive a simple and efficient algorithm, and also give evidence that its sample complexity is optimal, among efficient algorithms. Experiments on sentiment analysis demonstrate the efficacy of the method, both in terms of accuracy and interpretability. View details
    Sub-Nyquist Radar Systems: Temporal, Spectral and Spatial Compression
    Yonina C. Eldar
    IEEE Signal Processing Magazine(2018) (to appear)
    Preview abstract Conventional radar transmits electromagnetic waves towards the targets of interest. In between the outgoing pulses, the radar measures the signal reflected from the targets to determine their presence, range, velocity and other characteristics. Radar systems face multiple challenges, generating many trade-offs such as bandwidth versus range resolution and dwell time versus Doppler resolution. In MIMO radar, high resolution requires a large aperture and high number of antennas, increasing hardware and processing requirements. Recently, novel approaches in sampling theory and radar signal processing have been proposed to allow target detection and parameter recovery from samples obtained below the Nyquist rate. These techniques exploit the sparsity of the target scene in order to reduce the required number of samples, pulses and antennas, breaking the link between bandwidth, dwell time and number of antennas on the one hand and range, Doppler and azimuth resolution, respectively, on the other. This review introduces this so-called sub-Nyquist radar paradigm and describes the corresponding sampling and recovery algorithms, that leverage compressed sensing techniques to perform time and spatial compression. We focus on non radar imaging applications and survey many recent compressed radar systems. Our goal is to review the main impacts of compressed radar on parameter resolution as well as digital and analog complexity. The survey includes fast and slow time compression schemes as well as spatial compression approaches. We show that beyond substantial rate reduction, compression may also enable communication and radar spectrum sharing. Throughout the paper, we consider both theoretical and practical aspects of compressed radar, and present hardware prototype implementations, demonstrating real-time target parameter recovery from low rate samples in pulse-Doppler and MIMO radars. View details
    SUMMeR: Sub-Nyquist MIMO Radar
    David Cohen
    Yonina C. Eldar
    Alexander M. Haimovich
    IEEE Transactions on Signal Processing, 66(2018), pp. 4315 - 4330
    Preview abstract Multiple-input multiple-output (MIMO) radar exhibits several advantages with respect to the traditional radar array systems in terms of flexibility and performance. However, MIMO radar poses new challenges for both hardware design and digital processing. In particular, achieving high azimuth resolution requires a large number of transmit and receive antennas. In addition, digital processing is performed on samples of the received signal, from each transmitter to each receiver, at its Nyquist rate, which can be prohibitively large when high resolution is needed. Overcoming the rate bottleneck, sub-Nyquist sampling methods have been proposed that break the link between radar signal bandwidth and sampling rate. In this paper, we extend these methods to MIMO configurations and propose a sub-Nyquist MIMO radar (SUMMeR) system that performs both time and spatial compression. We present a range-azimuth-Doppler recovery algorithm from sub-Nyquist samples obtained from a reduced number of transmitters and receivers, that exploits the sparsity of the recovered targets' parameters. This allows us to achieve reduction in the number of deployed antennas and the number of samples per receiver, without degrading the time and spatial resolutions. Simulations illustrate the detection performance of SUMMeR for different compression levels and shows that both time and spatial resolution are preserved, with respect to classic Nyquist MIMO configurations. We also examine the impact of design parameters, such as antennas' locations and carrier frequencies, on the detection performance, and provide guidelines for their choice. View details
    Analog-to-Digital Cognitive Radio: Sampling, Detection, and Hardware
    Shahar Tsiper
    Yonina C. Eldar
    IEEE Signal Processing Magazine, 35(2018), pp. 137 - 166
    Preview abstract The radio spectrum is the radio-frequency (RF) portion of the electromagnetic spectrum. These spectral resources are traditionally allocated to licensed or primary users (PUs) by governmental organizations. As discussed in "Radio-Frequency Spectral Resources," most of the frequency bands are already allocated to one or more PUs. Consequently, new users cannot easily find free frequency bands. Spurred by the everincreasing demand from new wireless communication applications, this issue has become critical over the past few years. View details
    Expediting exploration by attribute-to-feature mapping for cold-start recommendations
    Michal Aharon
    Yair Koren
    Oren Somekh
    Raz Nissim
    RecSys '17 Proceedings of the Eleventh ACM Conference on Recommender Systems(2017), pp. 184-192
    Preview abstract The item cold-start problem is inherent to collaborative filtering (CF) recommenders where items and users are represented by vectors in a latent space. It emerges since CF recommenders rely solely on historical user interactions to characterize their item inventory. As a result, an effective serving of new and trendy items to users may be delayed until enough user feedback is received, thus, reducing both users' and content suppliers' satisfaction. To mitigate this problem, many commercial recommenders apply random exploration and devote a small portion of their traffic to explore new items and gather interactions from random users. Alternatively, content or context information is combined into the CF recommender, resulting in a hybrid system. Another hybrid approach is to learn a mapping between the item attribute space and the CF latent feature space, and use it to characterize the new items providing initial estimates for their latent vectors. In this paper, we adopt the attribute-to-feature mapping approach to expedite random exploration of new items and present LearnAROMA - an advanced algorithm for learning the mapping, previously proposed in the context of classification. In particular, LearnAROMA learns a Gaussian distribution over the mapping matrix. Numerical evaluation demonstrates that this learning technique achieves more accurate initial estimates than logistic regression methods. We then consider a random exploration setting, in which new items are further explored as user interactions arrive. To leverage the initial latent vector estimates with the incoming interactions, we propose DynamicBPR - an algorithm for updating the new item latent vectors without retraining the CF model. Numerical evaluation reveals that DynamicBPR achieves similar accuracy as a CF model trained on all the ratings, using 71% less exploring users than conventional random exploration. View details
    Sub-Nyquist cyclostationary detection for cognitive radio
    Yonina C. Eldar
    IEEE Transactions on Signal Processing, 65(2017), pp. 3004 - 3019
    Preview abstract Cognitive radio requires efficient and reliable spectrum sensing of wideband signals. In order to cope with the sampling rate bottleneck, new sampling methods have been proposed that sample below the Nyquist rate. However, such techniques decrease the signal-to-noise ratio (SNR), deteriorating the performance of subsequent energy detection. Cyclostationary detection, which exploits the periodic property of communication signal statistics, absent in stationary noise, is a natural candidate for this setting. In this paper, we consider cyclic spectrum recovery from sub-Nyquist samples, in order to achieve both efficiency and robustness to noise. To that end, we propose a structured compressed sensing algorithm, which extends orthogonal matching pursuit to account for the structure imposed by cyclostationarity. Next, we derive a lower bound on the sampling rate required for perfect cyclic spectrum recovery in the presence of stationary noise. In particular, we show that the cyclic spectrum can be reconstructed from samples obtained at 4/5 of the Nyquist rate, without any sparsity constraints on the signal. If the signal of interest is sparse, then the sampling rate may be further reduced to 8/5 of the Landau rate. Once the cyclic spectrum is recovered, we estimate the number of transmissions that compose the input signal, along with their carrier frequencies and bandwidths. Simulations show that cyclostationary detection outperforms energy detection in low SNRs in the sub-Nyquist regime. This was already known in the Nyquist regime, but is even more pronounced at sub-Nyquist sampling rates. View details
    Spectrum Sharing Radar: Coexistence via Xampling
    Kumar Vijay Mishra
    Yonina C. Eldar
    IEEE Transactions on Aerospace and Electronic Systems, 54(2017), pp. 1279 - 1296
    Preview abstract We present a Xampling-based technology enabling interference-free operation of radar and communication systems over a common spectrum. Our system uses a recently developed cognitive radio (CRo) to sense the spectrum at low sampling and processing rates. The Xampling-based cognitive radar (CRr) then transmits and receives in the available disjoint narrow bands. Our main contribution is the unification and adaptation of two previous ideas-CRo and CRr-to address spectrum sharing. Hardware implementation shows robust performance at SNRs up to -5 dB. View details
    CaSCADE: Compressed carrier and DOA estimation
    Shahar Stein Ioushua
    Or Yair
    Yonina C. Eldar
    IEEE Transactions on Signal Processing, 65(2017), pp. 2645 - 2658
    Preview abstract Spectrum sensing and direction of arrival (DOA) estimation have both been thoroughly investigated. Estimating the support of a set of signals and their DOAs is crucial to many signal processing applications, such as cognitive radio (CR). A challenging scenario, faced by CRs, is that of multiband signals, composed of several narrowband transmissions spread over a wide spectrum each with unknown carrier frequency and DOA. The Nyquist rate of such signals is high and constitutes a bottleneck for both analog and digital processing. To alleviate the sampling rate issue, several sub-Nyquist sampling methods, such as multicoset or the modulated wideband converter (MWC), have been proposed in the context of spectrum sensing. In this paper, we first suggest an alternative sub-Nyquist sampling and signal reconstruction method to the MWC, based on a uniform linear array (ULA). We then extend our approach to joint spectrum sensing and DOA estimation and propose the CompreSsed CArrier and DOA Estimation (CaSCADE) system, composed of an L-shaped array with two ULAs. In both cases, we derive conditions for perfect recovery of the signal parameters and the signal itself and provide two reconstruction algorithms. The first is based on the ESPRIT method and the second on compressed sensing techniques. Both our joint carriers and DOA recovery algorithms overcome the well-known pairing issue between the two parameters. Simulations demonstrate joint carrier and DOA recovery from CaSCADE sub-Nyquist samples. In addition, we show that our alternative spectrum sensing system outperforms the MWC in terms of recovery error and design complexity. View details
    Sub-Nyquist sampling for power spectrum sensing in cognitive radios: A unified approach
    Yonina C. Eldar
    IEEE Transactions on Signal Processing, 62(2014), pp. 3897-3910
    Preview abstract In light of the ever-increasing demand for new spectral bands and the underutilization of those already allocated, the concept of Cognitive Radio (CR) has emerged. Opportunistic users could exploit temporarily vacant bands after detecting the absence of activity of their owners. One of the crucial tasks in the CR cycle is therefore spectrum sensing and detection which has to be precise and efficient. Yet, CRs typically deal with wideband signals whose Nyquist rates are very high. In this paper, we propose to reconstruct the power spectrum of such signals from sub-Nyquist samples, rather than the signal itself as done in previous work, in order to perform detection. We consider both sparse and non sparse signals as well as blind and non blind detection in the sparse case. For each one of those scenarios, we derive the minimal sampling rate allowing perfect reconstruction of the signal’s power spectrum in a noise-free environment and provide power spectrum recovery techniques that achieve those rates. The analysis is performed for two different signal models considered in the literature, which we refer to as the analog and digital models, and shows that both lead to similar results. Simulations demonstrate power spectrum recovery at the minimal rate in noise-free settings and the impact of several parameters on the detector performance, including signal-to-noise ratio, sensing time and sampling rate. View details