הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

EE Seminar: Advancing Deep Learning by Integrating Signal Processing Techniques

24 בפברואר 2025, 12:00

אולם 011, בניין כיתות חשמל

EE Seminar: Advancing Deep Learning by Integrating Signal Processing Techniques

Electrical Engineering Systems Seminar

Speaker: Shady Abu-Hussein

Ph.D. student under the supervision of Prof. Raja Giryes

Monday, 24^th February 2025, at 12:00 (INSTEAD 19.2.25)

Room 011, Kitot Building, Faculty of Engineering

Advancing Deep Learning by Integrating Signal Processing Techniques

Abstract

Deep learning has revolutionized problem-solving across various domains, particularly in inverse problems. It enables robust reconstruction of signals or images from incomplete or noisy measurements. By leveraging vast amounts of data, these approaches have demonstrated significant improvements over traditional methods.

Generative modeling is another area that has gained prominence for its ability to synthesize realistic data by capturing the underlying distribution of input data. Techniques such as generative adversarial networks (GANs) and denoising diffusion probabilistic models (DDPMs) have paved the way for creating high-fidelity data. Through this research, we explore how one may advance deep learning approaches by integrating signal processing techniques, focusing on inversion tasks and generative models.

In the first part of the talk, we examine deep neural networks (DNNs) for inverse problems, particularly super-resolution tasks, where a high-resolution image is reconstructed from its low-resolution version. These models, however, struggle with unseen observation models. To address this, we introduce a correction filter from generalized sampling theory, enabling off-the-shelf super-resolvers to perform well across various observation models without retraining.

Next, we demonstrate how one may improve DDPMs by reducing their high computational costs, where high-quality image generation typically requires hundreds of large network evaluations. We generalize the forward diffusion process to include downsampling before adding noise perturbations. We show that using this approach, it is possible to generate images with 30% of the computations required in a single denoising diffusion step while obtaining significantly better generation results. Additionally, this approach improves the interpretability due to the fewer number of diffusion steps and smaller latent variables.

These advancements highlight the potential of integrating signal processing with deep learning to improve generalization in inverse problems and enhance the efficiency of generative models, paving the way for more practical and interpretable applications.

השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז. בדף הנוכחות שיועבר באולם במהלך הסמינר

EE Seminar: Leveraging Transfer Learning and NLP Techniques for Enhanced HTR of Ancient Jewish Manuscripts

23 במרץ 2025, 15:00

חדר 011, בניין כיתות חשמל

EE Seminar: Leveraging Transfer Learning and NLP Techniques for Enhanced HTR of Ancient Jewish Manuscripts

Electrical Engineering Systems Seminar

Speaker: Yoav Amiel

M.Sc. student under the supervision of Dr. Wasim Huleihel

Sunday, 2^nd February 2025, at 15:00

Room 011, Kitot Building, Faculty of Engineering

Direction-of-Arrival Estimation Using SubSpace Methods for Sparse Arrays

Abstract

Sparse arrays enable resolving more direction of arrivals (DoAs) than antenna elements using non-uniform arrays. This is typically achieved by reconstructing the covariance of a virtual large uniform linear array (ULA), which is then processed by subspace DoA estimators. However, these methods assume that the signals are non-coherent and the array is calibrated; the latter is often challenging to achieve in sparse arrays, where one cannot access the virtual array elements. In some real scenarios such as Track-Before-Detect (TBD), the receiver has no access to prior information on the sources, not even their number, which is typically needed or added as a hidden assumption to classic algorithms input. In this thesis, we propose Sparse-SubspaceNet, which leverages deep learning to enable subspace-based DoA recovery from sparse miscalibrated arrays with coherent sources without receiver prior information but for the number of sources, and suggest a scheme for learning the number of sources from covariance matrix eigenvalues distribution. Sparse-SubspaceNet utilizes a dedicated deep network to learn from data how to compute a surrogate virtual array covariance that is divisible into distinguishable subspaces. By doing so, we learn to cope with coherent sources and miscalibrated sparse arrays, while preserving the interpretability and the suitability of model-based subspace DoA estimators.

השתתפות בסמינר תיתן קרדיט שמיעה לתלמידי תואר שני ושלישי = עפ"י רישום שם מלא + מספר ת.ז. בדף הנוכחות שיועבר באולם במהלך הסמינר

סדנה של אולגה דגטיארבה- כיצד להתחיל לכתוב ולבצע את עבודת המחקר שלך

03 במרץ 2025, 14:00 - 15:00

Monday March 3th 2025 at 14:00

Wolfson Building of Mechanical Engineering, Room 206

Abstarct:

This is not a lecture: it’s a hands-on workshop! Bring your laptop and a long-overdue research paper, grant application, or PhD thesis, and let’s work on it together.

Many researchers struggle with writing, whether it’s a project that keeps getting postponed, a draft that has been forgotten for months, or something urgent with an approaching deadline that feels overwhelming. This workshop is designed specifically for PhD students, postdocs, and early-career researchers who need to break through procrastination and finally make progress.

We’ll tackle the challenges that keep you stuck, help you find your starting point, and guide you through re-animating an old draft or making a fresh start. You’ll leave with clarity, focus, and a realistic, doable plan to keep going after the session, so you can finally get your writing done in a short amount of time.

Join us, get unstuck, and take real action on your writing!

biosketch:

Olga is a Productivity Mentor at the Productivity for Scientists Ltd, based in the UK. Since 2010, she has been helping scientists overcome procrastination, boost productivity and take control of their priorities. Olga holds a PhD from the University of Edinburgh (2003) and has an extensive research background in high-pressure physics and crystallography. She is the recipient of an international prize for her contributions to high-pressure physics and has authored and co-authored 38 scientific papers, including Nature Materials, Nature Physics and Physical Review Letters.

סמינר מחלקתי של קרישנו קומר- גלי רוח צעירים בנוכחות זרם

24 בפברואר 2025, 14:00 - 15:00

Monday February 24th 2025 at 14:00

Wolfson Building of Mechanical Engineering, Room 206

Abstract:

The excitation mechanism of wind waves has been a topic of interest for more than a century; numerous experimental studies complemented by theoretical investigations have been conducted to understand this complex phenomenon. However, various aspects of this process remain unclear. The generation and evolution of wind waves are also affected by parameters such as direction and magnitude of wind, bathymetry, salinity, water current, etc. In the open ocean, wind waves often interact with currents driven by wind, tides, strong oceanic currents, etc., as well as with currents in confined environments such as in harbor entrances, river mouths, and lakes.

While the effect of uniform current on deep-water gravity waves has been extensively studied in both field and laboratory settings, however, only limited information is currently available on the diverse facets of the excitation of waves by wind that blows over flowing water. The present study aims to bridge that gap by providing a comprehensive set of experimental data under controlled laboratory setting in the absence and presence of current.

The experiments are carried out in the fully automated wind-wave flume for a wide range of wind and water velocities. The current is generated in both along-wind and counter -wind direction. Surface elevation and other wave parameters along the fetch are measured using both conventional analog and optical sensors. The turbulent velocity profile in air over random, three-dimensional moving water surface is measured using Pitot tube, while water velocity is measured using particle image velocimetry (PIV); and velocity at the air-water interface is determined using particle tracking velocimetry (PTV). The accumulated results are discussed in the framework of viscous shear flow instability theory.

Bio:

Krishanu Kumar earned his B. Tech from the School of Mechanical Engineering at SRM IST, India. In 2020, he joined the research group of Prof. Shemer as a MSc student, during which he collaborated with Prof. Barnea on studying interfacial waves in a confined environment. In 2021, he transferred to a direct PhD track under the supervision of Prof. Shemer. His research focuses on the excitation, interaction, and evolution of non-linear random wind waves in the presence of mean current.