Jump to main content
Begegnungszone: Statistical Physics and Machine Learning
Program
Begegnungszone: Statistical Physics and Machine Learning 

Statistical Physics and Machine Learning: Program

Please find below the preliminary program of the workshop as of September 15, 2023. You can also consult the more complete book of abstracts for more details on the presentations, including posters.

 
September 18, 2023 - Seminaris Hotel Leipzig
12:30-13:50 - Lunch at Seminaris Hotel -
13:50-14:00 - Opening -
Session I: Hidden Structures (Chair: Johannes Zierenberg)
14:00-14:30 Jeff Byers Statistical Mechanics and the Geometry of Machine Learning
14:45-15:00 Andrea Perin Quantifying the Improvements in Learning Efficiency Obtained by Leveraging Symmetries in the Data
15:15-15:30 NN Brainstorming session
15:45-16:15 - Coffee Break -
Session II: Interpretable Machine Learning (Chair: Riccardo Zecchina)
16:15-16:45 Marvin Wright Interpretable Machine Learning
17:00-17:15 Konstantin Nikolaou Towards a Phenomenological Understanding of Neural Networks: Data
17:30-17:45 NN Brainstorming session
18:30-20:00 - Dinner at Seminaris Hotel -
20:00-22:00 - Posters and welcome reception -
 
September 19, 2023 - Seminaris Hotel Leipzig
Session III: Deep Networks (Chair: Jonathan Kadmon)
08:30-09:00 Riccardo Zecchina Liquid States in Neural Networks
09:15-09:30 Francesco Cagnetta Learning Hierarchical Compositionality With Deep Convolutional Networks: Insights From a Random Hierarchy Model
09:45-10:00 Sebastiano Ariosto A Statistical Mechanics Framework for Deep Neural Networks Beyond the Infinite-Width Limit
10:00-10:30 - Coffee Break -
Session IV: Training I (Chair: Sebastian Goldt)
10:30-11:00 Beatriz Seoane Statistical Physics of Energy-Based Generative Models
11:15-11:30 Frederike Richert The Role of the Activation Function in Feedforward Learning Systems
11:45-12:00 NN Brainstorming session
12:20-12:30 - Conference photo -
12:30-14:00 - Lunch at Seminaris Hotel -
Session V: Training II (Chair: Beatriz Seoane)
14:00-14:30 Sebastian Goldt The Gaussian World is Not Enough -- How Training Data Shapes Neural Representations
14:45-15:00 Christian Keup Learning Dynamics in Deep Teacher-Student Models
15:15-15:30 NN Brainstorming session
15:45-16:15 - Coffee Break -
16:15-18:30 - Conference Excursion -
19:00-21:00 - Conference Dinner -
 
September 20, 2023 - Seminaris Hotel Leipzig
Session VI: Quantum Mechanics (Chair: Wolfhard Janke)
08:30-09:00 Lei Wang Unlocking the Power of the Variational Free-Energy Principle with Deep Generative Models
09:15-09:30 Luciano Vitteriti Transformer Variational Wave Functions for Frustrated Quantum Spin Systems
09:45-10:00 NN Brainstorming Session
10:00-10:30 - Coffee Break -
Session VII: Recurrent Networks (Chair: Marylou Gabrié)
10:30-11:00 Jonathan Kadmon Reliable Coding With Chaotic Neural Networks
11:15-11:30 Freya Behrens The Backtracking Dynamical Cavity Method and Cellular Automata
11:45-12:00 NN Brainstorming session
12:30-14:00 - Lunch at Seminaris Hotel -
Session VIII: Sampling (Chair: Jeff Byers)
14:00-14:30 Marylou Gabrié Assisting Sampling (of Physical States) With Generative Models
14:45-15:00 Henrik Christiansen Learning How to Integrate: Accelerating Hamiltonian Monte Carlo with Machine Learning
15:15-15:30 NN Brainstorming session
15:45-16:15 - Coffee Break -
16:15-18:30 - Group Projects -
18:30-20:00 - Dinner at Seminaris Hotel -
 
September 21, 2023 - Seminaris Hotel Leipzig
Session IX: Physical Systems (Chair: Martin Weigel)
08:30-09:00 Miriam Klopotek From Model Systems of Matter to Physical Computing and Physics-Explainable Machine Learning
09:15-09:30 Riccardo Rende Optimal Inference of a Generalised Potts Model by Single-Layer Transformers with Factored Attention
09:45-10:00 NN Brainstorming Session
10:00-10:30 - Coffee Break -
10:30-12:00 - Panel Discussion: Open Problems and Challenges -
12:00-12:15 - Summary -
12:15-12:30 - Conference Closing -
12:30-14:00 - Lunch at Seminaris Hotel -

Social Media

Connect with Us: