DNS: Determinantal Point Process Based Neural Network Sampler for Ensemble Reinforcement Learning

TL;DR

This paper introduces DNS, a neural network sampling method using determinantal point processes to efficiently select subsets of networks during training, reducing computation while maintaining or improving performance in reinforcement learning.

Contribution

The paper presents DNS, a novel DPP-based sampler that reduces training costs in ensemble reinforcement learning by selectively sampling networks for backpropagation.

Findings

DNS reduces training computation by over 50% in FLOPS.

DNS-augmented REDQ outperforms baseline REDQ in average reward.

The method is effective in continuous control tasks on MuJoCo environments.

Abstract

Application of ensemble of neural networks is becoming an imminent tool for advancing the state-of-the-art in deep reinforcement learning algorithms. However, training these large numbers of neural networks in the ensemble has an exceedingly high computation cost which may become a hindrance in training large-scale systems. In this paper, we propose DNS: a Determinantal Point Process based Neural Network Sampler that specifically uses k-dpp to sample a subset of neural networks for backpropagation at every training step thus significantly reducing the training time and computation cost. We integrated DNS in REDQ for continuous control tasks and evaluated on MuJoCo environments. Our experiments show that DNS augmented REDQ outperforms baseline REDQ in terms of average cumulative reward and achieves this using less than 50% computation when measured in FLOPS.