Integrated Calendar

Over the past decade, deep learning algorithms have achieved—and in some cases surpassed—human-level performance in face recognition. This remarkable success raises a fundamental question: to what extent do these artificial systems capture the mechanisms that underlie human face recognition? In this talk, I will explore the convergences and divergences between deep learning models and the human face recognition system. I will first show how deep convolutional neural networks (DCNNs)
reproduce key phenomena observed in human face perception. Yet, despite these similarities, important differences remain in how humans and deep learning algorithms learn and represent faces. To bridge these gaps, we employ models that learn continually and integrate visual and language-based models, to capture both perceptual and conceptual aspects of face recognition. Together, these findings demonstrate how deep learning algorithms can advance our understanding of human face recognition.

BIO:
Galit Yovel is a professor in the School of Psychological Sciences and Sagol School of Neuroscience at Tel Aviv University. She earned her PhD in Psychology from the University of Chicago and completed her Post-Doctoral studies in the Department of Brain and Cognitive Sciences at MIT. In her research she combines methods from experimental psychology, neuroimaging and AI to unravel the neural and cognitive mechanisms of human face recognition. Her work extends beyond faces to examine how the body, voice, motion, and semantic information contribute to person recognition. She was the head of Strauss MRI Center at Tel Aviv University (2015-2017), the head of the School of Psychological Sciences (2017-2021) and the head of the AI and Data Science major for students in life science/social sciences and law (2022-2025). She is the recipient of the Bruno award (2017), and a six-time recipient
of the Tel Aviv University Rector award for excellence in teaching.

Intrinsically disordered proteins (IDPs) and disordered protein regions, which comprise over 40% of the eukaryotic proteome, exhibit complex dynamics, fluctuating between diverse conformational ensembles. Unlike structured proteins, where short-range interactions and long-range contacts dictate singular three-dimensional folding, IDPs lack a single stable structure. To understand their biological function, it is crucial to establish a correlation between the amino acid sequence and the statistical properties of their structural ensemble.In this talk, I will present our recent work on neurofilament proteins, which are essential neuronal-specific cytoskeletal components containing large intrinsically disordered domains. Our study spans multiple length scales—from nanoscopic to macroscopic—aiming to uncover the molecular mechanisms underlying their functional behavior. By leveraging coarse-grained polymer physics models and integrating minimal parameters, we demonstrate that the structural ensemble of neurofilament proteins can be reasonably predicted. However, our findings underscore that specific sequence motifs and the surrounding context are necessary to fully capture the protein’s conformational landscape in solution.These results highlight the power of advanced polymer theories in describing the ensemble behavior of IDPs, offering a promising avenue for modeling their function and dysfunction, particularly in neurodegenerative disease contexts. By bridging the gap between sequence specificity and polymer physics, we aim to establish a more comprehensive framework for predicting IDP behavior and its implications in health and disease.Students interested in meeting the speaker after the seminar may sign up here:LINKFOR THE LATEST UPDATES AND CONTENT ON SOFT MATTER AND BIOLOGICAL PHYSICS AT THE WEIZMANN, VISIT OUR WEBSITE: https://www.bio

Symplectic capacities are invariants that quantify the size of symplectic manifolds using themes from Hamiltonian dynamics and symplectic topology. While convexity is not preserved under symplectomorphisms, convex domains nevertheless exhibit notable behavior with respect to these capacities. Viterbo's volume-capacity conjecture (2000) suggests that, among convex domains of equal volume, the ball has maximal capacity. By capturing the interplay between convex and symplectic geometries, this simply formulated conjecture has become highly influential in the study of symplectic capacities, prompting extensive research. One result in this direction shows that smooth domains which are symplectic Zoll—a dynamical property—are local maximizers. In this talk, I will present a counterexample to Viterbo’s conjecture developed jointly with Yaron Ostrover and discuss follow-up questions. One implication is that a capacity maximizer cannot be smooth and strictly convex, raising the question of characterizing nonsmooth dynamical properties that detect local maximizers. I will propose a dynamical extension of the Zoll property to nonsmooth domains and discuss its equivalence with certain topological properties.

Nonreciprocal interactions in which the influence of A on B differs from that of B on A are abundant in physical, chemical, biological, and ecological systems, and are known to give rise to oscillatory states. Yet, it remains unclear whether these states represent true phases of matter: Can they maintain long-range order in spatially extended, noisy environments in the thermodynamic limit? And what kinds of phase transitions do they exhibit? To address these questions, we introduce a minimal generalization of the Ising model with two species having opposing goals. We demonstrate that oscillatory phases are stable in three dimensions but not in two, and that nonreciprocity changes the critical exponents from those of the Ising model to those of the XY model. We further extend this framework to a nonreciprocal XY model and develop a Harris-like criterion that determines when nonreciprocity fundamentally alters universal behavior. Finally, we apply these insights to a recent model of biomolecular condensates, predicting exotic dynamical phases and suggesting experimental tests.

perovskites are among the most promising materials for next-generation solar cells, offering exceptionalefficiency gains and driving major investment in large-scale production. Yet, as the technology moves toward realworlddeployment, corrosion has emerged as a critical but often overlooked challenge. It arises not only fromenvironmental exposure but also from the intrinsic reactivity of the perovskite itself, which can attack metalelectrodes such as gold through complex chemical pathways.This show highlights why corrosion in perovskite devices is both subtle and important. Light and heat can triggerchemical changes that produce reactive species, either directly corroding metals or transforming the perovskite intoa more aggressive state. By connecting principles from corrosion science and semiconductor physics, we revealhow these reactions originate and what must be done to control them at their source.

Dear colleagues,Attached is the flyer for our Demonstration & Training Seminar: Advanced NEW Technologies for Extracellular Vesicle Research, taking place on December 11th, 2025, Benozio Building, 2nd floor.The session will feature two new platforms at WIS:ZetaView Nanoparticle Analyzer (now available at WIS) – particle concentration, size measurements, zeta potential and fluorescence-based phenotyping.Exodus Bio automated EV-isolation systems – high-purity, reproducible EV isolation with minimal hands-on time.You are welcome to join onsite or online:https://us02web.zoom.us/j/86837215413?pwd=iRTFVP4C2ykvJspMsZ8b2cJj8r5oJl.1Looking forward to seeing you there,Avi

A fundamental theorem states that a two-dimensional Riemannian manifold with boundary, equipped with a symmetric tensor field, is locally isometrically embedded in Euclidean space if and only if the symmetric tensor field satisfies the Gauss-Mainardi-Codazzi equations—in which case, the tensor field is the second fundamental form.

When the intrinsic metric is Euclidean, it is a classical result that such tensor fields are Hessians of functions satisfying the Monge-Ampère equation. I shall present a version of this result to arbitrary Riemannian metrics, using a generalized Hodge theory I developed for a broader class of geometric boundary-value problems. I will discuss this theory, its main features, and perhaps give a glimpse of more complicated examples it addresses.