A Simpler Alternative to Variational Regularized Counterfactual Risk Minimization

TL;DR

This paper proposes a simpler and more effective method for off-policy learning by directly minimizing f-divergence, improving upon the previous variational regularized approach.

Contribution

It introduces a novel direct approximation method for f-divergence minimization, replacing the more complex f-GAN based lower bound approach.

Findings

Direct divergence minimization outperforms f-GAN based methods

Reproducing original VRCRM results was unsuccessful

Proposed method shows empirical improvements in experiments

Abstract

Variance regularized counterfactual risk minimization (VRCRM) has been proposed as an alternative off-policy learning (OPL) method. VRCRM method uses a lower-bound on the $f$-divergence between the logging policy and the target policy as regularization during learning and was shown to improve performance over existing OPL alternatives on multi-label classification tasks. In this work, we revisit the original experimental setting of VRCRM and propose to minimize the $f$-divergence directly, instead of optimizing for the lower bound using a $f$-GAN approach. Surprisingly, we were unable to reproduce the results reported in the original setting. In response, we propose a novel simpler alternative to f-divergence optimization by minimizing a direct approximation of f-divergence directly, instead of a $f$-GAN based lower bound. Experiments showed that minimizing the divergence using $f$-GANs did not work as expected, whereas our proposed novel simpler alternative works better empirically.