arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.06313cs.LGcs.AI
收藏

面向深度强化学习的最优策略稀疏化与低秩分解

Optimal Policy Sparsification and Low Rank Decomposition for Deep Reinforcement Learning

Detroit Country Day School(底特律乡村日校)

Vikram Goddla

AI总结:

针对深度强化学习稠密策略算力消耗高、易过拟合、推理受限的问题,提出基于最优稀疏映射的$L_0$范数正则化技术,可在保证奖励不衰减的前提下实现策略稀疏化与低秩压缩,在多环境验证效果优异。

摘要:

深度强化学习(DRL)已在包括电脑游戏、机器人技术在内的广泛应用中展现出巨大潜力。然而,训练DRL策略会消耗海量计算资源,且得到的稠密策略易出现过拟合。此外,使用稠密DRL策略进行推理会限制其实际应用,尤其是在边缘计算场景中。\n剪枝、奇异值分解等技术已被应用于深度学习模型,以实现稀疏化与模型压缩,从而抑制过拟合、降低内存占用。但这类技术会导致性能次优,奖励出现明显衰减。已有研究提出$L_1$和$L_2$正则化技术用于神经网络稀疏化与稀疏自编码器开发,但它们在DRL环境中的落地应用尚不明确。\n我们提出一种基于最优稀疏映射的新型$L_0$范数正则化技术,用于对DRL策略进行稀疏化,同时推动策略分解为更低秩,且不会造成奖励衰减。我们在五种不同环境(Cartpole-v1、Acrobat-v1、LunarLander-v2、SuperMarioBros-7.1.v0以及外科手术机器人学习环境)中,使用多种同策略(on-policy)和异策略(off-policy)算法对所提$L_0$范数正则化技术进行了评估。\n实验结果表明,在SuperMarioBros环境中,经$L_0$范数正则化的DRL策略可实现93%的稀疏度,经低秩分解后可获得70%的压缩率,同时性能显著优于稠密策略。此外,在外科手术机器人学习环境中,经$L_0$范数正则化的DRL策略可实现36%的稀疏度,分解为低秩后可获得46%的压缩率,且仍保持良好性能。这些结果表明,我们提出的用于DRL策略稀疏化的定制化$L_0$范数正则化技术是减少计算资源消耗、抑制过拟合的一种极具前景的方案。

英文摘要:

Deep reinforcement learning(DRL) has shown significant promise in a wide range of applications including computer games and robotics. Yet, training DRL policies consume extraordinary computing resources resulting in dense policies which are prone to overfitting. Moreover, inference with dense DRL policies limit their practical applications, especially in edge computing. Techniques such as pruning and singular value decomposition have been used with deep learning models to achieve sparsification and model compression to limit overfitting and reduce memory consumption. However, these techniques resulted in sub-optimal performance with notable decay in rewards. $L_1$ and $L_2$ regularization techniques have been proposed for neural network sparsification and sparse auto-encoder development, but their implementation in DRL environments has not been apparent. We propose a novel $L_0$-norm-regularization technique using an optimal sparsity map to sparsify DRL policies and promote their decomposition to a lower rank without decay in rewards. We evaluated our $L_0$-norm-regularization technique across five different environments (Cartpole-v1, Acrobat-v1, LunarLander-v2, SuperMarioBros-7.1.v0 and Surgical Robot Learning) using several on-policy and off-policy algorithms. We demonstrated that the $L_0$-norm-regularized DRL policy in the SuperMarioBros environment achieved 93% sparsity and gained 70% compression when subjected to low-rank decomposition, while significantly outperforming the dense policy. Additionally, the $L_0$-norm-regularized DRL policy in the Surgical Robot Learning environment achieved a 36% sparsification and gained 46% compression when decomposed to a lower rank, while being performant. The results suggest that our custom $L_0$-norm-regularization technique for sparsification of DRL policies is a promising avenue to reduce computational resources and limit overfitting.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑