Distribution State

Distribution state, in the context of machine learning, refers to the challenge of handling data points that fall outside the distribution of training data (out-of-distribution or OOD). Current research focuses on developing robust models and algorithms that can either accurately classify or reject OOD samples, with a particular emphasis on techniques like attention mechanisms, energy-based loss functions, and contrastive learning for improved performance. This is crucial for building reliable AI systems in real-world applications, where encountering unseen data is inevitable, and for improving the safety and generalization capabilities of reinforcement learning agents. Addressing OOD issues is vital for advancing the trustworthiness and applicability of machine learning across diverse domains.

Papers

May 31, 2024

Skeleton-OOD: An End-to-End Skeleton-Based Model for Robust Out-of-Distribution Human Action Detection
Jing Xu, Anqi Zhu, Jingyu Lin, Qiuhong Ke, Cunjian Chen
Full Model End to End Distribution Detection Skeleton Based Action Recognition Human Action Recognition Action Detection Distribution State

December 10, 2023

Modifying RL Policies with Imagined Actions: How Predictable Policies Can Enable Users to Perform Novel Tasks
Isaac Sheidlower, Reuben Aronson, Elaine Short
Reinforcement Learning Robot Person Robot Teleoperation User Base Novel Task Distribution State

December 5, 2023

Colour versus Shape Goal Misgeneralization in Reinforcement Learning: A Case Study
Karolis Ramanauskas, Özgür Şimşek
Reinforcement Learning Strong Generalization Case Study Value Laden Choice Color Object Maze Environment Distribution State

November 3, 2023

AlberDICE: Addressing Out-Of-Distribution Joint Actions in Offline Multi-Agent RL via Alternating Stationary Distribution Correction Estimation
Daiki E. Matsunaga, Jongmin Lee, Jaeseok Yoon, Stefanos Leonardos, Pieter Abbeel, Kee-Eung Kim
Offline Reinforcement Learning DIstribution Estimation Offline Multi Agent Reinforcement Learning Distribution State

August 2, 2023

Achieving state-of-the-art performance in the Medical Out-of-Distribution (MOOD) challenge using plausible synthetic anomalies
Sergio Naval Marimont, Giacomo Tarroni
Distribution Detection State of the Art Challenge Task Unsupervised Anomaly Detection Music Mood Distribution State Synthetic Anomaly Anomaly Synthesis

June 22, 2023

Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning
Jinxin Liu, Ziqi Zhang, Zhenyu Wei, Zifeng Zhuang, Yachen Kang, Sibo Gai, Donglin Wang
Offline Reinforcement Learning Cross Domain Model Based Offline Reinforcement Learning Distribution State

March 28, 2023

Offline RL with No OOD Actions: In-Sample Learning via Implicit Value Regularization
Haoran Xu, Li Jiang, Jianxiong Li, Zhuoran Yang, Zhaoran Wang, Victor Wai Kin Chan, Xianyuan Zhan
Offline Reinforcement Learning Implicit Regularization Q$ Learning Distribution State

January 27, 2023

In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States
Fernando Castañeda, Haruki Nishimura, Rowan McAllister, Koushil Sreenath, Adrien Gaidon
Learning Based Control Barrier Function State Space Visuo Motor Control Distribution State Self Supervised Reinforcement Learning

December 19, 2022

Answering Complex Logical Queries on Knowledge Graphs via Query Computation Tree Optimization
Yushi Bai, Xin Lv, Juanzi Li, Lei Hou
Knowledge Graph Conjunctive Query Complex Query Complex Query Answering Distribution State Join Algorithm

November 15, 2022

Heatmap-based Out-of-Distribution Detection
Julia Hornauer, Vasileios Belagiannis
Distribution Detection Distribution Sample Distribution Image Distribution State

October 31, 2022

What is my math transformer doing? -- Three results on interpretability and generalization
François Charton
Strong Generalization Inherent Interpretability Key Result Distribution State Model Failure Arithmetic Transformer

October 22, 2022

OOD-DiskANN: Efficient and Scalable Graph ANNS for Out-of-Distribution Queries
Shikhar Jaiswal, Ravishankar Krishnaswamy, Ankit Garg, Harsha Vardhan Simhadri, Sheshansh Agrawal
High Efficiency Search Query Approximate Nearest Neighbor Search Efficient Query Distribution State Graph Based Approximate Nearest Neighbor

July 18, 2022

Back to the Manifold: Recovering from Out-of-Distribution States
Alfredo Reichlin, Giovanni Luca Marchetti, Hang Yin, Ali Ghadirzadeh, Danica Kragic
Distribution Shift Behavior Cloning Robot Action Robot Policy Dimensional Manifold Distribution State Distribution Shift Problem

June 21, 2022

Robust Task Representations for Offline Meta-Reinforcement Learning via Contrastive Learning
Haoqi Yuan, Zongqing Lu
Contrastive Learning Task Representation Offline Meta Reinforcement Learning Distribution State

Distribution State

Papers

Skeleton-OOD: An End-to-End Skeleton-Based Model for Robust Out-of-Distribution Human Action Detection

Modifying RL Policies with Imagined Actions: How Predictable Policies Can Enable Users to Perform Novel Tasks

Colour versus Shape Goal Misgeneralization in Reinforcement Learning: A Case Study

AlberDICE: Addressing Out-Of-Distribution Joint Actions in Offline Multi-Agent RL via Alternating Stationary Distribution Correction Estimation

Achieving state-of-the-art performance in the Medical Out-of-Distribution (MOOD) challenge using plausible synthetic anomalies

Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning

Offline RL with No OOD Actions: In-Sample Learning via Implicit Value Regularization

In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States

Answering Complex Logical Queries on Knowledge Graphs via Query Computation Tree Optimization

Heatmap-based Out-of-Distribution Detection

What is my math transformer doing? -- Three results on interpretability and generalization

OOD-DiskANN: Efficient and Scalable Graph ANNS for Out-of-Distribution Queries

Back to the Manifold: Recovering from Out-of-Distribution States

Robust Task Representations for Offline Meta-Reinforcement Learning via Contrastive Learning