Active Hypothesis Testing

Active hypothesis testing aims to efficiently determine the correct hypothesis among several possibilities by strategically selecting and acquiring data. Current research heavily utilizes deep learning, particularly reinforcement learning and neuroevolutionary algorithms, often within multi-agent frameworks to optimize data acquisition and decision-making, even in decentralized or unknown environments. These advancements improve the efficiency and accuracy of hypothesis testing across diverse applications, ranging from anomaly detection in sensor networks to two-sample testing in high-dimensional data analysis. The development of anytime-valid p-values and adaptive query strategies further enhances the practical applicability of these methods.

Papers

March 15, 2024

Single- and Multi-Agent Private Active Sensing: A Deep Neuroevolution Approach
George Stamatelis, Angelos-Nikolaos Kanatas, Ioannis Asprogerakas, George C. Alexandropoulos
Multi Agent Single Label Active Hypothesis Testing

September 14, 2023

Deep Multi-Agent Reinforcement Learning for Decentralized Active Hypothesis Testing
Hadar Szostak, Kobi Cohen
Multi Agent Reinforcement Learning Multi Agent Policy Single Agent Learning Active Hypothesis Testing

March 19, 2023

Active hypothesis testing in unknown environments using recurrent neural networks and model free reinforcement learning
George Stamatelis, Nicholas Kalouptsidis
Deep Reinforcement Learning Full Model Recurrent Neural Network Supervised Learning Unknown Environment Continuous Variable Active Hypothesis Testing

January 30, 2023

Active Sequential Two-Sample Testing
Weizhi Li, Karthikeyan Natesan Ramamurthy, Prad Kadambi, Pouria Saidi, Gautam Dasarathy, Visar Berisha
Log Likelihood Two Sample Active Hypothesis Testing

October 24, 2022

E-Valuating Classifier Two-Sample Tests
Teodora Pandeva, Tim Bakker, Christian A. Naesseth, Patrick Forré
Active Hypothesis Testing Classifier Two Sample Test

Active Hypothesis Testing

Papers

Single- and Multi-Agent Private Active Sensing: A Deep Neuroevolution Approach

Deep Multi-Agent Reinforcement Learning for Decentralized Active Hypothesis Testing

Active hypothesis testing in unknown environments using recurrent neural networks and model free reinforcement learning

Active Sequential Two-Sample Testing

E-Valuating Classifier Two-Sample Tests