לוח שנה משולב

In recent years, chemistry education has increasingly emphasized the integration of the Sustainable Development Goals (SDGs) into curricula, aiming to foster responsible global citizenship. In this lecture, we will explore two educational programs designed to enhance critical thinking, digital literacy, and student engagement with environmental issues. I will present findings on how these initiatives foster students’ critical thinking, encourage data-driven activism, and strengthen their sustainability agency. By showcasing the transformative potential of Education for Sustainable Development (ESD) in the chemistry classroom, I will highlight the role of science education in empowering students to drive meaningful change.

This talk presents two wavelet-based innovations that significantly improve convolutional neural networks (CNNs) in terms of scalability, computational efficiency, and robustness. I’ll begin with WTConv, a novel convolutional layer that leverages multilevel Haar wavelet decomposition to expand receptive fields efficiently. WTConv scales logarithmically with kernel size, enabling nearly global receptive fields without the parameter bloat typical of large kernels. Beyond improving classification accuracy, WTConv boosts shape bias and robustness to corruptions, all while remaining a lightweight, drop-in replacement for depthwise convolutions.

Next, I’ll introduce Wavelet Compressed Convolution (WCC), a method for compressing high-resolution activation maps in image-to-image tasks. By applying joint wavelet-domain shrinkage across channels and executing 1×1 convolutions directly on the compressed representation, WCC substantially reduces memory bandwidth and compute cost. Unlike aggressive quantization—which often causes severe degradation—WCC maintains high accuracy across tasks like segmentation, super-resolution, and depth estimation, even under extreme compression. Together, these methods show how wavelet transforms can serve as a powerful, hardware-friendly toolset for designing scalable and efficient CNNs.

Bio:
Shahaf Finder is a PhD candidate in Computer Science, supervised by Prof. Oren Freifeld and Prof. Eran Treister, researching efficient neural networks with a focus on convolutional neural networks (CNNs) and graph neural networks (GNNs). He is also a Co-Founder of LimitlessCNC, a startup focused on integrating AI into CNC programming, where he leads the development of the core algorithmic technology.

I will present the results of recent experiments with ribbons made of active, "BZ gel". The ribbons "flap", i.e.,   periodically change their curvature. When confined to a liquid interface,   the ribbons periodically "surf" from the center of the container to its walls.We analyze this motion and suggest that it represents a new, generic, type of locomotion; locomotion via curvature mismatch. In the experiments, the fluid interface is curved. When the curvatures of the ribbon and the surface are different,   both the ribbon and interface are deformed, a deformation that costs energy. Gradients of this energy lead to forces and torques on the ribbon and to its motion. We solve the equation of motion and successfully compute the trajectories of the active ribbons.Our model suggests that such motion could occur in purely solid systems. Specifically, it allows a flexible sheet, which is confined to curved (flexible or rigid) surface, to propagate without applying tangential forces. The possible relevance of this model to "curvotaxis", the phenomenon in which cells propagate and orient themselves in correlation with the substrate curvature, will be discussed. students whom are interested with a personal meeting, (The registration is limited to 5 participants) ( if you are interested in personal meeting please register)https://forms.gle/vfr9a2K3fYyKfx3d7 FOR THE LATEST UPDATES AND CONTENT ON SOFT MATTER AND BIOLOGICAL PHYSICS AT THE WEIZMANN, VISIT OUR WEBSITE: https://www.biosoftweizmann.com/

I will describe a joint work with Uri Bader, Michael Glasner and Roman Sauer in which we study unitary representations of semisimple Lie Groups that are cohomological; namely, such that the continuous cohomology of the group with coefficients in these representations is non-zero. This cohomology is a topological vector space that is not necessarily Hausdorff. The maximal Hausdorff quotient of it is called the reduced cohomology.

An essential work of Vogan and Zuckerman from 84’ solves the question completely for irreducible representations. From Vogan–Zuckerman's work one can also deduce a complete description of representations that have non-zero reduced cohomology: these are precisely representations that admit irreducible cohomological representations as sub-representations.

The question that is left is hence determining unitary representations that admit non-reduced cohomology. Up until now, non-reduced cohomology was mainly studied in the first degree, where its existence is equivalent to not having Kazhdan's property (T): H^1(G,V) is non-reduced if and only if V admits almost invariant vectors (that are not actually invariant). We prove some analoge (but more complicated) results in higher degrees, and settle the question of which representations admit non-reduced cohomology in terms of the Fell topology on the unitary dual of the group.