CoBERL: 强化学习的对比学习BERT

Jul, 2021

CoBERL: Contrastive BERT for Reinforcement Learning

Andrea Banino, Adrià Puidomenech Badia, Jacob Walker, Tim Scholtes, Jovana Mitrovic...

TL;DR该论文提出了通过使用新的对比度损失和混合LSTM-transformer架构组合来改进Reinforcement Learning(强化学习)代理的数据效率，从而实现在广泛领域内对像素的高效、稳健学习的Contrastive BERT for RL (CoBERL)。该方法通过双向掩码预测与最近对比方法的泛化相结合，学习RL中更好的Transformer表示，同时在Atari套装、控制任务和具有挑战性的3D环境中保持性能的一致改进。

Abstract

Many reinforcement learning (RL) agents require a large amount of experience to solve tasks. We propose contrastive bert for RL (CoBERL), an agent that combines a new contrastive loss and a hybrid LSTM-transforme