Combinatorial Generalization

Combinatorial generalization focuses on developing machine learning models capable of extrapolating to unseen combinations of features or inputs, going beyond simple interpolation within the training data distribution. Current research explores this challenge across various domains, including reinforcement learning, material science, and multi-agent systems, employing techniques like contrastive learning, successor features, and knowledge-aware parsimony learning to improve generalization performance. This ability to synthesize knowledge from diverse experiences holds significant promise for enhancing the efficiency and robustness of AI systems in complex, real-world scenarios, particularly where data scarcity or high dimensionality pose challenges.

Papers

June 29, 2024

Knowledge-Aware Parsimony Learning: A Perspective from Relational Graphs
Quanming Yao, Yongqi Zhang, Yaqing Wang, Nan Yin, James Kwok, Qiang Yang
Visual Perspective Relation Graph Combinatorial Generalization Parsimony Learning

June 24, 2024

Learning Temporal Distances: Contrastive Successor Features Can Provide a Metric Structure for Decision-Making
Vivek Myers, Chongyi Zheng, Anca Dragan, Sergey Levine, Benjamin Eysenbach
Reinforcement Learning Successor Feature Combinatorial Generalization Metric Structure Temporal Distance

January 20, 2024

Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of View
Raj Ghugare, Matthieu Geist, Glen Berseth, Benjamin Eysenbach
Reinforcement Learning Strong Generalization Supervised Learning Best View TD Learning Combinatorial Generalization

October 28, 2023

Towards Combinatorial Generalization for Catalysts: A Kohn-Sham Charge-Density Approach
Phillip Pope, David Jacobs
Catalyst Design Density Functional Theory Kohn Sham Combinatorial Generalization Charge Density

April 27, 2023

Learning to Extrapolate: A Transductive Approach
Aviv Netanyahu, Abhishek Gupta, Max Simchowitz, Kaiqing Zhang, Pulkit Agrawal
LeArning Abstract Reparameterization Method Transductive Approach Combinatorial Generalization Poor Extrapolation Performance

March 20, 2023

Neural Constraint Satisfaction: Hierarchical Abstraction for Combinatorial Generalization in Object Rearrangement
Michael Chang, Alyssa L. Dayan, Franziska Meier, Thomas L. Griffiths, Sergey Levine, Amy Zhang
Offline Reinforcement Learning Visual Input Object Rearrangement Constraint Learning Rearrangement Task Combinatorial Generalization Multilevel Abstraction

June 10, 2022

Large-Scale Retrieval for Reinforcement Learning
Peter C. Humphreys, Arthur Guez, Olivier Tieleman, Laurent Sifre, Théophane Weber, Timothy Lillicrap
Reinforcement Learning Deep Reinforcement Learning Large Scale Context Information Large Scale Retrieval Combinatorial Generalization

April 5, 2022

Lost in Latent Space: Disentangled Models and the Challenge of Combinatorial Generalisation
Milton L. Montero, Jeffrey S. Bowers, Rui Ponte Costa, Casimir J. H. Ludwig, Gaurav Malhotra
Generative Model Full Model Latent Space Challenge Task Generative Process Generative Factor Combinatorial Generalization

January 31, 2022

Generalization in Cooperative Multi-Agent Systems
Anuj Mahajan, Mikayel Samvelyan, Tarun Gupta, Benjamin Ellis, Mingfei Sun, Tim Rocktäschel, Shimon Whiteson
Strong Generalization Multi Agent Reinforcement Learning Multi Agent Cooperative Multi Agent System Combinatorial Generalization

Combinatorial Generalization

Papers

Knowledge-Aware Parsimony Learning: A Perspective from Relational Graphs

Learning Temporal Distances: Contrastive Successor Features Can Provide a Metric Structure for Decision-Making

Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of View

Towards Combinatorial Generalization for Catalysts: A Kohn-Sham Charge-Density Approach

Learning to Extrapolate: A Transductive Approach

Neural Constraint Satisfaction: Hierarchical Abstraction for Combinatorial Generalization in Object Rearrangement

Large-Scale Retrieval for Reinforcement Learning

Lost in Latent Space: Disentangled Models and the Challenge of Combinatorial Generalisation

Generalization in Cooperative Multi-Agent Systems