Random Network Distillation

Random Network Distillation (RND) is a technique used to enhance exploration in reinforcement learning by creating an intrinsic reward signal based on the difference between a randomly initialized network's predictions and a learned predictor network's predictions. Current research focuses on improving RND's robustness and effectiveness, addressing issues like unstable reward signals, inefficient exploration, and limitations in representation learning through methods such as pre-training and distributional approaches. These advancements are significant for improving the sample efficiency and performance of reinforcement learning agents in complex, sparse-reward environments, with applications ranging from robotics (e.g., AGV path planning) to data selection for neural network training.

Papers

November 4, 2024

Efficient Active Imitation Learning with Random Network Distillation
Emilien Biré, Anthony Kobanda, Ludovic Denoyer, Rémy Portelas
Locomotion Control Random Network Distillation Active Imitation

October 2, 2024

PreND: Enhancing Intrinsic Motivation in Reinforcement Learning through Pre-trained Network Distillation
Mohammadamin Davoodabadi, Negin Hashemi Dijujin, Mahdieh Soleymani Baghshah
Reinforcement Learning Intrinsic Reward Intrinsic Motivation Random Network Distillation

April 19, 2024

Random Network Distillation Based Deep Reinforcement Learning for AGV Path Planning
Huilin Yin, Shengkai Su, Yinjia Lin, Pengju Zhen, Karin Festl, Daniel Watzenig
Deep Reinforcement Learning Sparse Reward Environment Guided Vehicle Random Network Distillation

January 18, 2024

Exploration and Anti-Exploration with Distributional Random Network Distillation
Kai Yang, Jian Tao, Jiafei Lyu, Xiu Li
Deep Reinforcement Learning Environment Exploration Efficient Exploration Random Network Distillation

May 1, 2023

Towards a Phenomenological Understanding of Neural Networks: Data
Samuel Tovey, Sven Krippendorf, Konstantin Nikolaou, Christian Holm
Neural Network Raw Data Data Selection Collective Variable Empirical Neural Tangent Kernel Random Network Distillation

March 16, 2023

Knowledge Distillation for Adaptive MRI Prostate Segmentation Based on Limit-Trained Multi-Teacher Models
Eddardaa Ben Loussaief, Hatem Rashwan, Mohammed Ayad, Mohammed Zakaria Hassan, Domenec Puig
Knowledge Distillation Deep Model Teacher Model Prostate Segmentation Random Network Distillation

March 2, 2023

Expert-Free Online Transfer Learning in Multi-Agent Reinforcement Learning
Alberto Castagna, Ivana Dusparic
Reinforcement Learning Multi Agent Reinforcement Learning Multi Agent Random Network Distillation

January 31, 2023

Anti-Exploration by Random Network Distillation
Alexander Nikulin, Vladislav Kurenkov, Denis Tarasov, Sergey Kolesnikov
Offline Reinforcement Learning Distribution Data Soft Actor Critic Sampling Free Random Network Distillation

October 5, 2022

On Neural Consolidation for Transfer in Reinforcement Learning
Valentin Guillet, Dennis G. Wilson, Carlos Aguilar-Melchor, Emmanuel Rachelson
Reinforcement Learning Deep Reinforcement Learning Transfer Learning Knowledge Transfer Formality Transfer Memory Consolidation Optimal Transport Distillation Random Network Distillation

May 13, 2022

Knowledge Distillation Meets Open-Set Semi-Supervised Learning
Jing Yang, Xiatian Zhu, Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
Knowledge Distillation Open Set Knowledge Representation Random Network Distillation