t-DGR：一种用于决策中持续学习的基于轨迹的深度生成重放方法
t-DGR: A Trajectory-Based Deep Generative Replay Method for Continual Learning in Decision Making
AI总结：
针对决策持续学习中自回归生成重放的轨迹误差累积问题，提出以轨迹时间步为条件的非自回归 t-DGR，并在 Continual World 平均成功率上达到最先进水平。
摘要：
深度生成重放已成为决策任务中持续学习的一种有前景的方法。该方法通过利用从先前遇到的任务生成轨迹来增强当前数据集，从而解决灾难性遗忘问题。然而，现有的用于持续学习的深度生成重放方法依赖于自回归模型，这类模型在生成的轨迹中会受到误差累积的影响。本文提出一种简单、可扩展且非自回归的方法，用于决策任务中的持续学习；该方法使用一个以轨迹时间步为条件来生成任务样本的生成模型。我们在 Continual World 基准上评估了该方法，结果表明，在持续学习方法的平均成功率指标上，我们的方法取得了最先进的性能。代码可在 https://github.com/WilliamYue37/t-DGR 获取。
英文摘要：
Deep generative replay has emerged as a promising approach for continual learning in decision-making tasks. This approach addresses the problem of catastrophic forgetting by leveraging the generation of trajectories from previously encountered tasks to augment the current dataset. However, existing deep generative replay methods for continual learning rely on autoregressive models, which suffer from compounding errors in the generated trajectories. In this paper, we propose a simple, scalable, and non-autoregressive method for continual learning in decision-making tasks using a generative model that generates task samples conditioned on the trajectory timestep. We evaluate our method on Continual World benchmarks and find that our approach achieves state-of-the-art performance on the average success rate metric among continual learning methods. Code is available at https://github.com/WilliamYue37/t-DGR.
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。
