בית הספר
להנדסת חשמל ומחשבים

Name: Shadi Kandalaft. Degree: Masters in Electrical and Computer Engineering. Supervisor: Prof. Adrian Stern. Subject: Optimization of hyperspectral imaging for precision medicine in cancer diagnostics. Abstract: This research focuses on optimizing hyperspectral imaging for cancer diagnostics through Fourier-based spectral imaging with Sagnac interferometry — a robust common-path interferometric method enabling rapid acquisition of interferograms from stained biopsies. By integrating deep learning and compressive sensing via learned partial transform ensembles (LPTnet), the system jointly learns optimal sampling patterns and reconstruction mappings. This approach aims to reduce the number of required interferogram samples, significantly accelerating image acquisition and lowering data size, while preserving high diagnostic accuracy. The study contributes to advancing rapid, information-rich, and computationally efficient hyperspectral imaging for precision cancer diagnostics. Abstract: Seminar Title: Study lineage development of cellular differentiation and maturation using single-cell RNA-seq: Optimization of computational methods of RNA kinetics to understand hematopoietic stem and progenitor cells in healthy, sick, and recovered mice. Pinkhas Likhterov M.Sc. Electrical Engineer Student, Ben-Gurion University of the Negev Abstract: The proposed research aims to elucidate the intricate processes governing hematopoietic stem and progenitor cell (HSC) differentiation and maturation, utilizing advanced single-cell RNA sequencing (scRNA-seq) techniques and computational methods. The study focuses on understanding the dynamic changes in HSCs in different physiological states: healthy, sick, and recovered mice. Chronic illnesses, such as Salmonella infection, cause significant alterations in HSC phenotypes, skewing their differentiation pathways towards specific lineages. This research uses deep-learning computational methods to decode these changes and their reversion upon recovery at single-cell resolution. Aim 1 constructs a comprehensive HSC atlas for each physiological state using scRNA-seq data. Aim 2 identifies differentially represented cell types, states, genes, and pathways across conditions, revealing mechanisms of disease response and recovery. Aim 3 optimizes RNA kinetics methods for scRNA-seq analysis, benchmarking RNA velocity approaches and improving them with deep learning frameworks. We further aim to implement a wrapper that integrates and reconciles the knowledge and congruence of all methods, providing a unified and more accurate view of dynamic cellular processes.

Title: Correcting Deletions: Fundamental Questions and Surprising Applications Abstract: Error-correcting codes are central to information theory and theoretical computer science, enabling reliable communication in the presence of noise. Classical results focus primarily on substitutions and erasures, and over the years elegant constructions have emerged that meet, or closely approach, the optimal rate–noise tradeoffs in these settings. A natural and equally fundamental error model involves synchronization errors, such as insertions and deletions. These errors, already studied in the 1960s, cause misalignment between sender and receiver and arise in various modern technologies, with DNA-based data storage being a particularly compelling example. Despite decades of attention, fully characterizing the capacity of synchronization channels and constructing practical, near-optimal codes for them remain major open challenges. In this talk, I will survey recent progress on coding for synchronization channels and then focus on the performance of linear and Reed–Solomon codes under insertions and deletions, demonstrating that well-structured algebraic codes can, perhaps unexpectedly, correct a significant amount of synchronization noise. I will then highlight a surprising application of these ideas in secret sharing. Finally, I will discuss how input-correlated insertion/deletion channels naturally arise in DNA-based data storage, an emerging ultra-dense archival technology, and present capacity theorems and efficient coding schemes tailored to an important class of such channels. Short Bio: Roni Con is a postdoctoral researcher at the Technion, hosted by Prof. Eitan Yaakobi. In Spring 2024, he was a Simons Research Fellow in the program Error-Correcting Codes: Theory and Practice at the Simons Institute for the Theory of Computing. He completed his Ph.D. in October 2023 at Tel Aviv University under the supervision of Profs. Amir Shpilka and Zachi Tamo. His research focuses on error-correcting codes and information theory, with applications to synchronization-error channels, modern storage systems, and cryptography.

Abstract: We seek algorithms for program learning that are both sample-efficient and computationally feasible. Classical results show that targets admitting short program descriptions (e.g., with short "python code") can be learned with a "small" number of examples (scaling with the size of the code) via length-first program enumeration, but the search is exponential in description length. Consequently, Gradient-based training avoids this cost yet can require exponentially many samples on certain short-program families. To address this gap, we introduce a propose-and-verify framework that replaces exhaustive enumeration with an LLM-guided search over candidate programs while retaining ERM-style selection on held-out data. Specifically, we draw k candidates with a pretrained reasoning-augmented LLM, compile and check each on the data, and return the best verified hypothesis, with no feedback, adaptivity, or gradients. Theoretically, we show that coordinate-wise online mini-batch SGD requires many samples to learn certain short programs. Empirically, our method solves tasks such as parity variants, pattern matching, and primality testing with as few as 200 samples, while SGD-trained transformers overfit even with 100,000 samples. These results indicate that language-guided program synthesis recovers much of the statistical efficiency of finite-class ERM while remaining computationally tractable, offering a practical route to learning succinct hypotheses beyond the reach of gradient-based training. Short bio: Tomer Galanti is an Assistant Professor of Computer Science and Engineering at Texas A&M University. His research combines theoretical and empirical approaches to understand the foundations of modern deep learning and large language models. Before joining Texas A&M, he was a Postdoctoral Associate at MIT’s Center for Brains, Minds and Machines (CBMM) and CSAIL, working with Tomaso Poggio, and a Research Scientist Intern at Google DeepMind. He received his Ph.D. in Computer Science from Tel Aviv University under the supervision of Lior Wolf. His work has appeared in leading venues such as NeurIPS, ICML, ICLR, and JMLR, and has been recognized with a NeurIPS oral presentation (2020) and an ICLR spotlight presentation (2025).

Name: Shadi Kandalaft. Degree: Masters in Electrical and Computer Engineering. Supervisor: Prof. Adrian Stern. Subject: Optimization of hyperspectral imaging for precision medicine in cancer diagnostics. Abstract: This research focuses on optimizing hyperspectral imaging for cancer diagnostics through Fourier-based spectral imaging with Sagnac interferometry — a robust common-path interferometric method enabling rapid acquisition of interferograms from stained biopsies. By integrating deep learning and compressive sensing via learned partial transform ensembles (LPTnet), the system jointly learns optimal sampling patterns and reconstruction mappings. This approach aims to reduce the number of required interferogram samples, significantly accelerating image acquisition and lowering data size, while preserving high diagnostic accuracy. The study contributes to advancing rapid, information-rich, and computationally efficient hyperspectral imaging for precision cancer diagnostics. name : Pinkhas Rafael Likhterov, an Electrical Engineer student at Ben-Gurion University of the Negev, supervised by Dr. Ofir Cohen (co-supervisor: Prof. Dan Vilenchik). Seminar Title: Study lineage development of cellular differentiation and maturation using single-cell RNA-seq: Optimization of computational methods of RNA kinetics to understand hematopoietic stem and progenitor cells in healthy, sick, and recovered mice. Abstract: The proposed research aims to elucidate the intricate processes governing hematopoietic stem and progenitor cell (HSC) differentiation and maturation, utilizing advanced single-cell RNA sequencing (scRNA-seq) techniques and computational methods. The study focuses on understanding the dynamic changes in HSCs in different physiological states: healthy, sick, and recovered mice. Chronic illnesses, such as Salmonella infection, cause significant alterations in HSC phenotypes, skewing their differentiation pathways towards specific lineages. This research uses deep-learning computational methods to decode these changes and their reversion upon recovery at single-cell resolution. Aim 1 constructs a comprehensive HSC atlas for each physiological state using scRNA-seq data. Aim 2 identifies differentially represented cell types, states, genes, and pathways across conditions, revealing mechanisms of disease response and recovery. Aim 3 optimizes RNA kinetics methods for scRNA-seq analysis, benchmarking RNA velocity approaches and improving them with deep learning frameworks. We further aim to implement a wrapper that integrates and reconciles the knowledge and congruence of all methods, providing a unified and more accurate view of dynamic cellular processes.

name : Noam Klainer Seminar subject: Estimating Expected Costs of DNA Synthesis Under Homopolymer Constraints Seminar Summary: My research focuses on estimating the expected costs associated with DNA synthesis under homopolymer constraints, which are of practical significance in the field of DNA-based data storage. The project is based on probabilistic modeling and draws on concepts from information theory to derive quantitative insights into constrained sequence generation and its associated costs. I am studying Ms.c on Electrical and Computer Engineering Supervisor: Dr Ohad Elishko. Name: Saar Ben-Yochana Master’s student in Communication Systems Engineering Supervisors: Prof. Chen Avin and Dr. Gabriel Scalosub "Seminar Title: “T3P: Topology-Tailored Tensor Parallel Abstract: "As deep learning models continue to grow in scale and complexity, methods of distributed machine learning training, and particularly those used for large language models (LLMs), have become a critical ingredient in making such computations efficient and feasible. In such contexts, tensor parallelism (TP) is widely employed to distribute computations across multiple accelerators. However, since TP mandates frequent and high-volume communication between devices, the underlying network characteristics significantly influence performance. Previous work was mostly either model-agnostic or topology-agnostic and did not pick provably optimal configurations. This study presents Topology-Tailored Tensor Parallelism, T3P, an efficient algorithm that identifies the communication-optimal TP sharding configuration (within the considered search space) based on both the model architecture and the network topology. In particular, we show that T3P is optimal for any given resharding cost model.”

Name : Amaljith Chandroth Kalliyadan Degree : PhD Name of the Supervisor : Professor. Ibrahim Abdulhalim Topic: Tunable metamaterial photonic structures for electro-optic and energy saving devices Summary Metamaterial (MTM) devices hold significant potential across various fields, including sensing, security, screening, and telecommunications. Liquid crystals (LCs) are particularly promising for tuning these devices due to their strong electro-optic response and ability to infiltrate nano- and micro-scale structures. Previous studies have demonstrated that metamaterial-based systems can achieve achromatic performance, wide viewing angles, and rapid switching, as only thin LC layers are required. In this research, I will report on several designs and some built devices: (i) A Thick dielectric subwavelength structure filled with LC showing that the switching speed increases by two orders of magnitude due to designing the electromagnetic field confinement in a region where the LC switches faster; (ii) A metal-insulator-metal (MIM) stacked MTM device for interfacial solar photothermal conversion. The structure comprises a top MTM metal layer for wide-angle light trapping, a middle dielectric spacer for resonant mode enhancement, and a bottom reflective tungsten layer to nullify transmission. It exhibits broadband, wide-angle absorption and polarization-independent characteristics, as well as a remarkable photothermal conversion efficiency in the 400-2400nm spectral range. An average absorption of ≈77% is theoretically and experimentally achieved. The device has shown a promising photothermal conversion efficiency of 73%, an evaporation rate (ER) above 10 L/h.m2 under moderate sunlight conditions, with the potential to double under optimized conditions, showing that it is a strong candidate for efficient water desalination and wastewater treatment; (iii) An enhanced optical transmission (EOT) structure combined with LC layer for fast and broadband tuning. Ofir Yaish PhD Prof. Yaron Orenstein and Dr. Nir Shlezinger subject: Computational modeling and analysis of biological systems based on high-throughput data Summary : The rapid growth of high-throughput experimental assays has opened unprecedented opportunities for understanding complex biological systems, but it has also introduced new challenges in data modeling and interpretation. In this seminar, I will present computational frameworks I developed for modeling and analyzing genomic systems, focusing on two distinct but complementary directions. First, I will discuss our work on CRISPR/Cas9 off-target prediction, where we generated the most comprehensive datasets of off-target sites with bulges to date and developed SWOffinder, a combinatorial search algorithm, alongside advanced machine-learning models that achieved state-of-the-art prediction accuracy across both in vitro and in cellula data. This study highlights the power of integrating large-scale data with tailored algorithms to address limitations in genome-editing safety and design. However, predictive modeling alone is insufficient for fully uncovering biological mechanisms. Therefore, in the second part, I will present F-MoDA (Fourier-based Motif Discovery in Attribution maps), a novel computational method for extracting regulatory motifs from deep-learning attribution maps. F-MoDA leverages signal processing and clustering to discover motifs more accurately, concisely, and efficiently than existing approaches. Together, these works demonstrate how computational modeling, combined with interpretable machine learning, can both enhance practical applications, such as genome editing, and advance our understanding of fundamental biological processes.

Name: Danielle Yaffe Degree: Electrical Engineering, Audio Signal Processing Professor: Prof. Boaz Rafaely Subject: “Audio-Visual Speech Enhancement for Spatial Audio” Abstract: Audio-visual speech enhancement (AVSE) has been found to be particularly useful at low signal-to-noise (SNR) ratios due to the immunity of the visual features to acoustic noise. However, a significant gap exists in AVSE methods tailored to enhance spatial audio under low-SNR conditions. The latter is of growing interest with augmented reality applications. To address this gap, we present a multi-channel AVSE framework based on VisualVoice that leverages spatial cues from microphone arrays and visual information for enhancing the target speaker in noisy environments. We also introduce MAVe, a novel database containing multi-channel audio-visual signals in controlled, reproducible room conditions across a wide range of SNR levels. Experiments demonstrate that the proposed method consistently achieves significant gains in SI-SDR, STOI, and PESQ, particularly in low SNRs. Binaural signal analysis further confirms the preservation of spatial cues and intelligibility. Abstract: Accurate state estimation is central to reliable power system operation. Classical weighted least-squares methods perform well under nominal conditions but are vulnerable to certain measurement anomalies. In particular, measurement sign errors can arise from hardware faults, communication issues, or malicious interference, and may reverse the physical meaning of measurements while remaining difficult to detect using conventional techniques. In this seminar, we present an approach that explicitly accounts for sign errors in power system measurements. By jointly estimating the system state and the underlying measurement signs, the proposed method enables the detection and correction of sign-flipped data that standard methods often miss. Simulation results on benchmark power systems demonstrate improved estimation accuracy and enhanced anomaly detection compared to traditional state estimation approaches. Name: Tslil Sardam Supervisors: Prof. Tirza Routtenberg and Prof. Eran Treister Degree: M.Sc. in Electrical & Computer Engineering (with Thesis) Seminar Topic: State Estimation with Measurement Sign Errors Abstract: Accurate state estimation is central to reliable power system operation. Classical weighted least-squares methods perform well under nominal conditions but are vulnerable to certain measurement anomalies. In particular, measurement sign errors can arise from hardware faults, communication issues, or malicious interference, and may reverse the physical meaning of measurements while remaining difficult to detect using conventional techniques. In this seminar, we present an approach that explicitly accounts for sign errors in power system measurements. By jointly estimating the system state and the underlying measurement signs, the proposed method enables the detection and correction of sign-flipped data that standard methods often miss. Simulation results on benchmark power systems demonstrate improved estimation accuracy and enhanced anomaly detection compared to traditional state estimation approaches