Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples

Abstract

The problem of detecting whether a test sample is from in-distribution (i.e., training distribution by a classifier) or out-of-distribution sufficiently different from it arises in many real-world machine learning applications. However, the state-of-art deep neural networks are known to be highly overconfident in their predictions, i.e., do not distinguish in- and out-of-distributions. Recently, to handle this issue, several threshold-based detectors have been proposed given pre-trained neural classifiers. However, the performance of prior works highly depends on how to train the classifiers since they only focus on improving inference procedures. In this paper, we develop a novel training method for classifiers so that such inference algorithms can work better. In particular, we suggest two additional terms added to the original loss (e.g., cross entropy). The first one forces samples from out-of-distribution less confident by the classifier and the second one is for (implicitly) generating most effective training samples for the first one. In essence, our method jointly trains both classification and generative neural networks for out-of-distribution. We demonstrate its effectiveness using deep convolutional neural networks on various popular image datasets.

References in corpus (5)

• Improving neural networks by preventing co-adaptation of feature detectors
• Striving for Simplicity: The All Convolutional Net
• On Calibration of Modern Neural Networks
• One weird trick for parallelizing convolutional neural networks
• Energy-based Generative Adversarial Network

Cited by in corpus (43)

• The Fishyscapes Benchmark: Measuring Blind Spots in Semantic Segmentation
• Measuring Calibration in Deep Learning
• The Conditional Entropy Bottleneck
• Testing for Outliers with Conformal p-values
• Unified Probabilistic Deep Continual Learning through Generative Replay and Open Set Recognition
• AI Research Considerations for Human Existential Safety (ARCHES)
• On the Validity of Bayesian Neural Networks for Uncertainty Estimation
• Enhancing the Robustness of Deep Neural Networks by Boundary Conditional GAN
• SoK: Machine Learning Governance
• A Unified Benchmark for the Unknown Detection Capability of Deep Neural Networks
• Open Set Medical Diagnosis
• Are all outliers alike? On Understanding the Diversity of Outliers for Detecting OODs
• One Versus all for deep Neural Network Incertitude (OVNNI) quantification
• Hybrid Models for Open Set Recognition
• Revisiting the Evaluation of Uncertainty Estimation and Its Application to Explore Model Complexity-Uncertainty Trade-Off
• Controlling Over-generalization and its Effect on Adversarial Examples Generation and Detection
• On the Role of Dataset Quality and Heterogeneity in Model Confidence
• Reliable and Trustworthy Machine Learning for Health Using Dataset Shift Detection
• Domain segmentation and adjustment for generalized zero-shot learning
• Pixel-wise Energy-biased Abstention Learning for Anomaly Segmentation on Complex Urban Driving Scenes
• $k$Folden: $k$-Fold Ensemble for Out-Of-Distribution Detection
• Few-Shot Open-Set Recognition using Meta-Learning
• Modeling Token-level Uncertainty to Learn Unknown Concepts in SLU via Calibrated Dirichlet Prior RNN
• Self-Supervised Representation Learning for Visual Anomaly Detection
• Toward Metrics for Differentiating Out-of-Distribution Sets
• Bigeminal Priors Variational auto-encoder
• Fine-grained Uncertainty Modeling in Neural Networks
• Utilizing Network Properties to Detect Erroneous Inputs
• Improving Classifier Confidence using Lossy Label-Invariant Transformations
• Robust Deep Learning Ensemble against Deception
• Removing Undesirable Feature Contributions Using Out-of-Distribution Data
• Cascade Watchdog: A Multi-tiered Adversarial Guard for Outlier Detection
• Improved Robustness to Open Set Inputs via Tempered Mixup
• Out-of-Scope Intent Detection with Self-Supervision and Discriminative Training
• Pixel Invisibility: Detecting Objects Invisible in Color Images
• Probabilistic Trust Intervals for Out of Distribution Detection
• Learning to Separate Clusters of Adversarial Representations for Robust Adversarial Detection
• OvA-INN: Continual Learning with Invertible Neural Networks
• Energy-based Unknown Intent Detection with Data Manipulation
• Joint Distribution across Representation Space for Out-of-Distribution Detection
• Understanding Classifier Mistakes with Generative Models
• Improve Uncertainty Estimation for Unknown Classes in Bayesian Neural Networks with Semi-Supervised /One Set Classification
• Improving robustness of classifiers by training against live traffic