Unbiased Decisions Reduce Regret: Adversarial Optimism for the Bank Loan Problem

Elena Gal, Shaun Singh, Aldo Pacchiano, Benjamin Walker, Terry Lyons, Jakob Nicolaus Foerster

Keywords: adversarial domain adaptation, online learning, adversarial de-biasing, neural optimism

TL;DR: We use adversarial domain adaptation to combat accumulating bias for a class of online learning problems.

Abstract:

In many real world settings binary classification decisions are made based on limited data in near real-time, e.g. when assessing a loan application. We focus on a class of these problems that share a common feature that the true label is only observed when a data point is assigned a positive label by a learner, e.g. we only learn of an outcome of \emph{accepted} loan applications. In this setting, sometimes referred to as the Bank Loan Problem (BLP) in the literature, the labelled training set suffers from accumulating bias since it is created by learners past decisions.
Prior work mitigates the consequences of this bias by injecting optimism into the model to allow the learner to correct self-reinforcing false rejections. This reduces long term regret but comes at the cost of a higher false acceptance rate.
We introduce \emph{adversarial optimism} (AdOpt) to directly address the bias in the training set using \emph{adversarial domain adaptation}. The goal of AdOpt is to learn an unbiased but informative representation of past data, by reducing the distributional shift between the set of \textit{accepted} data points and all data points seen thus far. AdOpt integrates classification made using this debiased representation of the data with the recently proposed \emph{pseudo-label optimism}(PLOT) method to increase the rate of correct decisions at every timestep.
AdOpt significantly exceeds state-of-the-art performance on a set of challenging BLP benchmark problems.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning