[Paper Review] Sample Efficient Actor-Critic with Experience Replay

Ziyu Wang, Victor Bapst+5 more|arXiv (Cornell University)|Nov 3, 2016

Reinforcement Learning in Robotics26 references221 citations

TL;DR

The provided text does not contain the actual paper content, only formatting instructions for NIPS 2016, so a summary cannot be produced.

ABSTRACT

This paper presents an actor-critic deep reinforcement learning agent with experience replay that is stable, sample efficient, and performs remarkably well on challenging environments, including the discrete 57-game Atari domain and several continuous control problems. To achieve this, the paper introduces several innovations, including truncated importance sampling with bias correction, stochastic dueling network architectures, and a new trust region policy optimization method.

Motivation & Objective

• No content about research objectives is provided in the supplied text.

Proposed method

• No content about proposed methods or key techniques is provided in the supplied text.

Experimental results

Research questions

• RQ1 No research questions are available in the supplied text.

Key findings

• No results or findings are available in the supplied text.