Towards Consistent Predictive Confidence through Fitted Ensembles

TL;DR

This paper proposes a new framework for evaluating classifier performance with out-of-distribution examples and introduces fitted ensembles, which improve predictive confidence and OOD detection without auxiliary data, demonstrated on multiple datasets.

Contribution

It introduces the separable concept learning framework for realistic performance measurement and proposes fitted ensembles as a novel method for consistent confidence and OOD detection.

Findings

Fitted ensembles outperform conventional ensembles on OOD detection.

The framework does not require auxiliary OOD datasets.

Fitted ensembles scale well to large datasets like ImageNet.

Abstract

Deep neural networks are behind many of the recent successes in machine learning applications. However, these models can produce overconfident decisions while encountering out-of-distribution (OOD) examples or making a wrong prediction. This inconsistent predictive confidence limits the integration of independently-trained learning models into a larger system. This paper introduces separable concept learning framework to realistically measure the performance of classifiers in presence of OOD examples. In this setup, several instances of a classifier are trained on different parts of a partition of the set of classes. Later, the performance of the combination of these models is evaluated on a separate test set. Unlike current OOD detection techniques, this framework does not require auxiliary OOD datasets and does not separate classification from detection performance. Furthermore, we present a new strong baseline for more consistent predictive confidence in deep models, called fitted ensembles, where overconfident predictions are rectified by transformed versions of the original classification task. Fitted ensembles can naturally detect OOD examples without requiring auxiliary data by observing contradicting predictions among its components. Experiments on MNIST, SVHN, CIFAR-10/100, and ImageNet show fitted ensemble significantly outperform conventional ensembles on OOD examples and are possible to scale.