Deep Gamblers: Learning to Abstain with Portfolio Theory

Abstract

We deal with the \textit{selective classification} problem (supervised-learning problem with a rejection option), where we want to achieve the best performance at a certain level of coverage of the data. We transform the original $m$-class classification problem to $(m+1)$-class where the $(m+1)$-th class represents the model abstaining from making a prediction due to disconfidence. Inspired by portfolio theory, we propose a loss function for the selective classification problem based on the doubling rate of gambling. Minimizing this loss function corresponds naturally to maximizing the return of a \textit{horse race}, where a player aims to balance between betting on an outcome (making a prediction) when confident and reserving one's winnings (abstaining) when not confident. This loss function allows us to train neural networks and characterize the disconfidence of prediction in an end-to-end fashion. In comparison with previous methods, our method requires almost no modification to the model inference algorithm or model architecture. Experiments show that our method can identify uncertainty in data points, and achieves strong results on SVHN and CIFAR10 at various coverages of the data.

arXiv commentCamera-Ready version for NeurIPS2019. Link to our code updated

References in corpus (22)

• Adam: A Method for Stochastic Optimization
• ADADELTA: An Adaptive Learning Rate Method
• An overview of gradient descent optimization algorithms
• Improved Regularization of Convolutional Neural Networks with Cutout
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Densely Connected Convolutional Networks
• Wide Residual Networks
• On the Convergence of Adam and Beyond
• Weight Uncertainty in Neural Networks
• Searching for Exotic Particles in High-Energy Physics with Deep Learning
• The Marginal Value of Adaptive Gradient Methods in Machine Learning
• Large-Margin Softmax Loss for Convolutional Neural Networks
• Deep Learning is Robust to Massive Label Noise
• Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing their Input Gradients
• Bayesian Recurrent Neural Networks
• One-Shot Federated Learning
• Neural Machine Translation and Sequence-to-sequence Models: A Tutorial
• Evidence for the Appearance of Atmospheric Tau Neutrinos in Super-Kamiokande
• A Walk with SGD
• Theory of Deep Learning III: explaining the non-overfitting puzzle
• Identity Crisis: Memorization and Generalization under Extreme Overparameterization
• Cautious Deep Learning

Cited by in corpus (2)

• Energy-based Detection of Adverse Weather Effects in LiDAR Data
• Learning Not to Learn in the Presence of Noisy Labels