Distance-based Confidence Score for Neural Network Classifiers

TL;DR

This paper introduces a scalable, distance-based confidence score for neural network classifiers that improves prediction reliability across various tasks by leveraging data embeddings from the penultimate layer.

Contribution

It proposes a novel, computationally efficient confidence scoring method using data embeddings, enhanced by distance-based loss or adversarial training, outperforming traditional scores.

Findings

Significant improvement in error prediction accuracy

Enhanced ensemble weighting performance

Better novelty detection results

Abstract

The reliable measurement of confidence in classifiers' predictions is very important for many applications and is, therefore, an important part of classifier design. Yet, although deep learning has received tremendous attention in recent years, not much progress has been made in quantifying the prediction confidence of neural network classifiers. Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with prohibitive computational costs. In this paper we propose a simple, scalable method to achieve a reliable confidence score, based on the data embedding derived from the penultimate layer of the network. We investigate two ways to achieve desirable embeddings, by using either a distance-based loss or Adversarial Training. We then test the benefits of our method when used for classification error prediction, weighting an ensemble of classifiers, and novelty detection. In all tasks we show significant improvement over traditional, commonly used confidence scores.