Learning to Repeat: Fine Grained Action Repetition for Deep\n Reinforcement Learning

2017/02/20 by Sahil Sharma, Sharma, Sahil, Balaraman Ravindran +3 · 6 citations
Computer Science · Engineering · #Advanced Memory and Neural Computing #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Artificial Intelligence in Games #FOS: Computer and information sciences #Ferroelectric and Negative Capacitance Devices #Machine Learning (cs.LG) #Neural and Evolutionary Computing (cs.NE) #Reinforcement Learning in Robotics

paper · pdf · doi:10.48550/arxiv.1702.06054

openalex publication_date 2017/02/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Reinforcement Learning algorithms can learn complex behavioral patterns for\nsequential decision making tasks wherein an agent interacts with an environment\nand acquires feedback in the form of rewards sampled from it. Traditionally,\nsuch algorithms make decisions, i.e., select actions to execute, at every\nsingle time step of the agent-environment interactions. In this paper, we\npropose a novel framework, Fine Grained Action Repetition (FiGAR), which\nenables the agent to decide the action as well as the time scale of repeating\nit. FiGAR can be used for improving any Deep Reinforcement Learning algorithm\nwhich maintains an explicit policy estimate by enabling temporal abstractions\nin the action space. We empirically demonstrate the efficacy of our framework\nby showing performance improvements on top of three policy search algorithms in\ndifferent domains: Asynchronous Advantage Actor Critic in the Atari 2600\ndomain, Trust Region Policy Optimization in Mujoco domain and Deep\nDeterministic Policy Gradients in the TORCS car racing domain.\n

Cited by

• TARC: Time-Adaptive Robotic Control
• Toward Agents That Reason About Their Computation
• Learning When to Plan: Efficiently Allocating Test-Time Compute for LLM Agents
• Learning to Mix n-Step Returns: Generalizing lambda-Returns for Deep Reinforcement Learning
• The Atari Grand Challenge Dataset
• Learning to Multi-Task by Active Sampling

Related

• State-Novelty Guided Action Persistence in Deep Reinforcement Learning
• Select before Act: Spatially Decoupled Action Repetition for Continuous Control
• Enhancing Deep Deterministic Policy Gradients on Continuous Control Tasks with Decoupled Prioritized Experience Replay
• Reinforcement Learning with Simple Sequence Priors
• Deep Reinforcement Learning with Adjustments