Learning Dynamic

Learning dynamics research investigates how models evolve during training, aiming to understand and optimize the process of acquiring knowledge and skills. Current efforts focus on characterizing learning trajectories across various model architectures, including neural networks (both deep and shallow), and algorithms like stochastic gradient descent, analyzing the influence of hyperparameters and initialization strategies on performance and stability. This research is crucial for improving model efficiency, generalization, and robustness in diverse applications, from robotics and reinforcement learning to understanding fundamental aspects of neural network behavior and biological learning.

Papers

October 5, 2022

The Influence of Learning Rule on Representation Dynamics in Wide Neural Networks
Blake Bordelon, Cengiz Pehlevan
External Influence Learning Dynamic Wide Neural Network Dynamic Representation Dynamic Prediction Learning Rule Infinite Width Neural Network

August 22, 2022

A simple learning agent interacting with an agent-based market model
Matthew Dicks, Andrew Paskaramoorthy, Tim Gebbie
Agent Smith Learning Dynamic Agent Based Model Multi Agent Market Optimal Execution

July 25, 2022

Live in the Moment: Learning Dynamics Model Adapted to Evolving Policy
Xiyao Wang, Wichayaporn Wongkamjan, Furong Huang
Model Based Reinforcement Learning Policy Learning Learning Dynamic Temporal Moment Learned Dynamic Mixture Policy Evolving Policy

July 21, 2022

The Neural Race Reduction: Dynamics of Abstraction in Gated Networks
Andrew M. Saxe, Shagun Sodhani, Sam Lewallen
Deep Learning Neural Architecture Learning Dynamic Cognitive Abstraction Gated Linear Structured Network

July 20, 2022

Learning Underspecified Models
In-Koo Cho, Jonathan Libgober
Learning Dynamic Optimal Decision Demand Curve Optimal Pricing

July 5, 2022

An Empirical Study of Implicit Regularization in Deep Offline RL
Caglar Gulcehre, Srivatsan Srinivasan, Jakub Sygnowski, Georg Ostrovski, Mehrdad Farajtabar, Matt Hoffman, Razvan Pascanu, Arnaud Doucet
Gradient Descent Offline Reinforcement Learning Empirical Study Learning Dynamic Stable Rank Implicit Regularization

July 1, 2022

A Theoretical Analysis of the Learning Dynamics under Class Imbalance
Emanuele Francazi, Marco Baity-Jesi, Aurelien Lucchi
Class Imbalance Theoretical Analysis Learning Dynamic Data Imbalance Sample Gradient

June 17, 2022

On the Influence of Enforcing Model Identifiability on Learning dynamics of Gaussian Mixture Models
Pascal Mattia Esser, Frank Nielsen
External Influence Gaussian Mixture Model Learning Dynamic Partial Identifiability Model Space Singular Learning

June 12, 2022

Convergence and Stability of Coupled Belief--Strategy Learning Dynamics in Continuous Games
Manxi Wu, Saurabh Amin, Asuman Ozdaglar
Early Stage Convergence Core Stability Learning Dynamic Belief State Strategic Agent Continuous Game Strategy Training

June 5, 2022

Learning Dynamics and Generalization in Reinforcement Learning
Clare Lyle, Mark Rowland, Will Dabney, Marta Kwiatkowska, Yarin Gal
Reinforcement Learning Strong Generalization Policy Gradient Learning Dynamic Temporal Difference Learning Deep Reinforcement Learning Agent

May 29, 2022

Independent and Decentralized Learning in Markov Potential Games
Chinmay Maheshwari, Manxi Wu, Druv Pai, Shankar Sastry
Reinforcement Learning Nash Equilibrium Decentralized Learning Learning Dynamic Markov Potential Game

May 19, 2022

Mean-Field Analysis of Two-Layer Neural Networks: Global Optimality with Linear Convergence Rates
Jingwei Zhang, Xunpeng Huang, Jincheng Yu
Mean Field Learning Dynamic Two Layer Neural Network Global Optimality Linear Convergence Rate Noisy Gradient Descent

April 25, 2022

Uncoupled Learning Dynamics with $O(\log T)$ Swap Regret in Multiplayer Games
Ioannis Anagnostides, Gabriele Farina, Christian Kroer, Chung-Wei Lee, Haipeng Luo, Tuomas Sandholm
Learning Dynamic Upper Bound General Sum Game Multiplayer Game Swap Regret

February 22, 2022

Learning Dynamics and Structure of Complex Systems Using Graph Neural Networks
Zhe Li, Andreas S. Tolias, Xaq Pitkow
Graph Neural Network Strong Generalization Inner Structure Learning Dynamic Complex System Belief Propagation Novel Generalization Probabilistic Inference Algorithm

February 16, 2022

The learning phases in NN: From Fitting the Majority to Fitting a Few
Johannes Schneider
Deep Neural Network Information Bottleneck Learning Dynamic Classification Error Layer Neural Network Majority Rule Learning Phase

January 21, 2022

Deep Q-learning: a robust control approach
Balazs Varga, Balazs Kulcsar, Morteza Haghir Chehreghani
Reinforcement Learning Learning Dynamic Deep Q Learning Robust Control Robust Controller Double Deep Q Learning

January 1, 2022

Distributed Evolution Strategies Using TPUs for Meta-Learning
Alex Sheng, Derek He
Meta Learning Learning Dynamic Evolution Strategy Tensor Processing Unit

December 2, 2021

Trap of Feature Diversity in the Learning of MLPs
Dongrui Liu, Shaobo Wang, Jie Ren, Kangrui Wang, Sheng Yin, Huiqi Deng, Quanshi Zhang
LeArning Abstract Multi Layer Learning Dynamic Multi Layer Perceptron Feature Diversity Field Trap

November 13, 2021