Predicting Classification Accuracy When Adding New Unobserved Classes

TL;DR

This paper introduces a method to predict the accuracy of multiclass classifiers on unseen classes by leveraging a novel measure called reversed ROC, enabling better extrapolation from limited class samples.

Contribution

The authors propose a new measure, rROC, and a neural network-based algorithm, CleaneX, for accurately estimating classifier accuracy on large, unobserved class sets.

Findings

CleaneX outperforms existing methods in accuracy prediction.

The rROC measure effectively relates classifier scores to accuracy.

Method validated on object detection, face recognition, and brain decoding datasets.

Abstract

Multiclass classifiers are often designed and evaluated only on a sample from the classes on which they will eventually be applied. Hence, their final accuracy remains unknown. In this work we study how a classifier's performance over the initial class sample can be used to extrapolate its expected accuracy on a larger, unobserved set of classes. For this, we define a measure of separation between correct and incorrect classes that is independent of the number of classes: the "reversed ROC" (rROC), which is obtained by replacing the roles of classes and data-points in the common ROC. We show that the classification accuracy is a function of the rROC in multiclass classifiers, for which the learned representation of data from the initial class sample remains unchanged when new classes are added. Using these results we formulate a robust neural-network-based algorithm, "CleaneX", which learns to estimate the accuracy of such classifiers on arbitrarily large sets of classes. Unlike previous methods, our method uses both the observed accuracies of the classifier and densities of classification scores, and therefore achieves remarkably better predictions than current state-of-the-art methods on both simulations and real datasets of object detection, face recognition, and brain decoding.