Abstract

Topic 1: Noise2Noise: Learning Image Restoration without Clean Data Overview: A simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption.

Topic 2: RL + Theoretical Particle Physics: Building Theories is Like Playing Chess Overview: I will discuss a broad overview of how reinforcement learning is being used to design new theories in particle physics, targeting open questions that the Standard Model cannot fully explain.

Papers that will be discussed/mentioned:

- Quark Mass Models and Reinforcement Learning [2103.04759] - Exploring the flavor structure of quarks and leptons with reinforcement learning [2304.14176] - Reinforcement learning-based statistical search strategy for an axion model from flavor [2409.10023] - Towards Beyond Standard Model Model-Building with Reinforcement Learning on Graphs [2407.07184, 2407.07203] - Towards AI-assisted Neutrino Flavor Theory Design [2506.08080]

It is not necessary to read the above literature before the session!