Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Behavior-Induced Mirror-Prox Temporal-Difference Learning for Faster Off-Policy Prediction

About

Gradient temporal-difference methods provide stable off-policy prediction with linear function approximation, but their practical performance is strongly affected by the geometry induced by the auxiliary-variable metric. Existing Mirror-Prox TD methods typically use the feature covariance metric, whereas hybrid TD methods suggest that behavior-policy transition information can provide a more informative update geometry. This paper proposes a behavior-induced Mirror-Prox temporal-difference method, called STHTD-MP, which replaces the covariance metric in the primal-dual saddle-point formulation with the symmetric part of the behavior-policy Bellman matrix. The method keeps a single learning rate for the primal and auxiliary variables and applies a Mirror-Prox prediction-correction step to the resulting hybrid saddle-point operator. We provide a formal convergence analysis for fixed-policy linear prediction under standard stochastic approximation assumptions: the behavior-induced metric is positive definite, the joint mean system is Hurwitz, boundedness follows from a Lyapunov argument, and the stochastic recursion converges by the ODE method. We further derive projected-oracle ergodic gap bounds and an exact mean-operator comparison with GTD2-MP based on the spectral radius of the deterministic Mirror-Prox error matrix. The analysis shows that STHTD-MP can have a smaller mean contraction factor than GTD2-MP when the behavior-induced metric improves the saddle-point geometry. Exact numerical mean-operator analysis on two-state, Random Walk, and Boyan Chain benchmarks supports this condition, while Baird's counterexample is identified as a singular boundary case where the strict assumptions fail.

Xingguo Chen, Yuchen Shen, Shangdong Yang, Chao Li, Guang Yang, Wenhao Wang• 2026

Related benchmarks

TaskDatasetResultRank
Off-policy predictionBoyan chain
Tail-average RMSE0.1687
16
Off-policy predictionBaird's counterexample
Steady-state AUC Error1.946
15
Off-policy predictionRandom Walk
Steady-state AUC Error0.0401
15
Off-policy predictionRandom Walk
RMSVE0.0394
9
Off-policy predictionTwo-state environment
Steady-state AUC Error6.71
9
Off-policy predictionTwo-state
RMSVE3.97
9
Off-policy predictionBoyan Chain environment
Steady-state AUC Error0.1692
9
Showing 7 of 7 rows

Other info

Follow for update