Intentional TD and Intentional Policy Gradient select step sizes for fixed fractional TD error reduction and bounded policy KL divergence, yielding stable streaming deep RL performance on par with batch methods.
https://arxiv.org/abs/2410.14606
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
representative citing papers
A generalization of differential TD extends it to episodic settings while preserving policy ordering, inheriting linear TD guarantees, and improving sample efficiency.
IRRL combines residual RL over a Social Force base with buffer-free incremental updates to match buffer-based social navigation performance and enable on-device real-world adaptation.
C51 matches StreamQ in streaming RL on 55 Atari games while a new Adaptive Q(λ) algorithm based on bounded derivatives and variance-adjusted updates reaches nearly double the human baseline.
citing papers explorer
-
Intentional Updates for Streaming Reinforcement Learning
Intentional TD and Intentional Policy Gradient select step sizes for fixed fractional TD error reduction and bounded policy KL divergence, yielding stable streaming deep RL performance on par with batch methods.
-
Extending Differential Temporal Difference Methods for Episodic Problems
A generalization of differential TD extends it to episodic settings while preserving policy ordering, inheriting linear TD guarantees, and improving sample efficiency.
-
Incremental Residual Reinforcement Learning Toward Real-World Learning for Social Navigation
IRRL combines residual RL over a Social Force base with buffer-free incremental updates to match buffer-based social navigation performance and enable on-device real-world adaptation.
-
Revisiting Adam for Streaming Reinforcement Learning
C51 matches StreamQ in streaming RL on 55 Atari games while a new Adaptive Q(λ) algorithm based on bounded derivatives and variance-adjusted updates reaches nearly double the human baseline.
- The Hive Mind is a Single Reinforcement Learning Agent