Latent Exploration

Latent exploration focuses on improving exploration strategies in reinforcement learning and other optimization problems by leveraging latent representations of the state or action space. Current research explores methods like perturbing latent variables within neural networks (e.g., using structured noise in policy networks or variational autoencoders) to guide exploration more effectively than traditional unstructured approaches. This addresses challenges like the curse of dimensionality and inefficient exploration in high-dimensional environments, leading to improved performance in tasks ranging from robotics control to game playing. The resulting advancements promise more efficient and robust learning algorithms for complex systems.

Papers

July 18, 2024

Random Latent Exploration for Deep Reinforcement Learning
Srinath Mahankali, Zhang-Wei Hong, Ayush Sekhari, Alexander Rakhlin, Pulkit Agrawal
Deep Reinforcement Learning Stochastic Reward Latent Exploration

June 14, 2024

ScaLES: Scalable Latent Exploration Score for Pre-Trained Generative Networks
Omer Ronen, Ahmed Imtiaz Humayun, Randall Balestriero, Richard Baraniuk, Bin Yu
Latent Space Variational Autoencoder Pre Trained Generative Model Latent Space Optimization Continuous Optimization Ghanaian Musical Scale Latent Exploration

May 31, 2023

Latent Exploration for Reinforcement Learning
Alberto Silvio Chiappa, Alessandro Marin Vargas, Ann Zixiang Huang, Alexander Mathis
Reinforcement Learning Pure Exploration State Perturbation Action Noise Latent Exploration

August 31, 2022

Cell-Free Latent Go-Explore
Quentin Gallouédec, Emmanuel Dellandréa
Reinforcement Learning Exploration Problem Go Explore Latent Exploration

Latent Exploration

Papers

Random Latent Exploration for Deep Reinforcement Learning

ScaLES: Scalable Latent Exploration Score for Pre-Trained Generative Networks

Latent Exploration for Reinforcement Learning

Cell-Free Latent Go-Explore