GradientDICE: rethinking generalized offline estimation of stationary values

Conference item

We present GradientDICE for estimating the density ratio between the state distribution of the target policy and the sampling distribution in off-policy reinforcement learning. GradientDICE fixes several problems of GenDICE (Zhang et al., 2020), the current state-of-the-art for estimating such density ratios. Namely, the optimization problem in GenDICE is not a convex-concave saddle-point problem once nonlinearity in optimization variable parameterization is introduced to ensure positivity, so primal-dual algorithms are not guaranteed to find the desired solution. However, such nonlinearity is essential to ensure the consistency of GenDICE even with a tabular representation. This is a fundamental contradiction, resulting from GenDICE’s original formulation of the optimization problem. In GradientDICE, we optimize a different objective from GenDICE by using the Perron-Frobenius theorem and eliminating GenDICE’s use of divergence, such that nonlinearity in parameterization is not necessary for GradientDICE, which is provably convergent under linear function approximation.

Authors: Zhang, S; Liu, B; Whiteson, S

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Journal of Machine Learning Research
• Host title: International Conference on Machine Learning, 13-18 July 2020, Virtual
• Series: Proceedings of Machine Learning Research
• Series number: 119
• Publication date: 2020-11-21
• Acceptance date: 2020-06-01
• Event title: 37th International Conference on Machine Learning (ICML 2020)
• Event location: Virtual
• Event website: https://icml.cc/Conferences/2020
• Event start date: 2020-07-12
• Event end date: 2020-07-18
• ISSN: 2640-3498
• Language: English
• Keywords: FFR