In this paper we introduce the idea of improving the performance of parametric temporal-difference (TD) learning algorithms by selectively emphasizing or de-emphasizing their updates on different time steps. In particular, we show that varying the emphasis of linear TD($\lambda$)'s updates in a particular way causes its expected update to become stable under off-policy training. The only prior model-free TD methods to achieve this with per-step computation linear in the number of function approximation parameters are the gradient-TD family of methods including TDC, GTD($\lambda$), and GQ($\lambda$). Compared to these methods, our _emphatic TD($\lambda$)_ is simpler and easier to use; it has only one learned parameter vector and one step-size parameter. On the other hand, the range of problems for which it is stable but does not converge with probability one is larger than for gradient-TD methods. Our treatment includes general state-dependent discounting and bootstrapping functions, and a way of specifying varying degrees of interest in accurately valuing different states.

本文介绍了通过有选择地加强或减弱不同时间步骤的更新来改善参数化时序差分学习算法的性能，并且提出了一种新的增强TD(λ)算法，相较于之前的方法，它更简单易用，只有一个学习参数向量和一个步长参数，含有状态相关折扣函数和自举函数，并且能够根据需要确定不同状态的精确定价程度。

一种强调的离线时序差分学习方法