Distance-based Confidence Score for Neural Network Classifiers

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.

Cited by in corpus (26)

• The Fishyscapes Benchmark: Measuring Blind Spots in Semantic Segmentation
• Why is the Mahalanobis Distance Effective for Anomaly Detection?
• Safer Classification by Synthesis
• Into the Unknown: Active Monitoring of Neural Networks
• Internet Congestion Control via Deep Reinforcement Learning
• Learning Densities in Feature Space for Reliable Segmentation of Indoor Scenes
• Hyperparameter-Free Out-of-Distribution Detection Using Softmax of Scaled Cosine Similarity
• The Hidden Uncertainty in a Neural Networks Activations
• Revisiting the Evaluation of Uncertainty Estimation and Its Application to Explore Model Complexity-Uncertainty Trade-Off
• Classification Confidence Estimation with Test-Time Data-Augmentation
• Controlling Risk of Web Question Answering
• A Simple and Effective Self-Supervised Contrastive Learning Framework for Aspect Detection
• Mitigating Uncertainty in Document Classification
• Experimental online quantum dots charge autotuning using neural networks
• Robust quantum dots charge autotuning using neural network uncertainty
• Explainable Artificial Intelligence (XAI) for Increasing User Trust in Deep Reinforcement Learning Driven Autonomous Systems
• Prediction Confidence from Neighbors
• Noise-Resistant Deep Metric Learning with Probabilistic Instance Filtering
• nn-dependability-kit: Engineering Neural Networks for Safety-Critical Autonomous Driving Systems
• DIRA: Dynamic Domain Incremental Regularised Adaptation
• Hack The Box: Fooling Deep Learning Abstraction-Based Monitors
• Relevant-features based Auxiliary Cells for Energy Efficient Detection of Natural Errors
• Leveraging Semantic Embeddings for Safety-Critical Applications
• Active Learning for UAV-based Semantic Mapping
• Out-of-Distribution Example Detection in Deep Neural Networks using Distance to Modelled Embedding
• Trust but Verify: Assigning Prediction Credibility by Counterfactual Constrained Learning