Deep Verifier Networks: Verification of Deep Discriminative Models with Deep Generative Models

Abstract

AI Safety is a major concern in many deep learning applications such as autonomous driving. Given a trained deep learning model, an important natural problem is how to reliably verify the model's prediction. In this paper, we propose a novel framework -- deep verifier networks (DVN) to verify the inputs and outputs of deep discriminative models with deep generative models. Our proposed model is based on conditional variational auto-encoders with disentanglement constraints. We give both intuitive and theoretical justifications of the model. Our verifier network is trained independently with the prediction model, which eliminates the need of retraining the verifier network for a new model. We test the verifier network on out-of-distribution detection and adversarial example detection problems, as well as anomaly detection problems in structured prediction tasks such as image caption generation. We achieve state-of-the-art results in all of these problems.

Accepted to AAAI 2021

References in corpus (27)

• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Deep Residual Learning for Image Recognition
• Generative Adversarial Text to Image Synthesis
• Conditional Image Generation with PixelCNN Decoders
• Densely Connected Convolutional Networks
• Wide Residual Networks
• On Calibration of Modern Neural Networks
• Learning deep representations by mutual information estimation and maximization
• A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks
• LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
• NIPS 2016 Tutorial: Generative Adversarial Networks
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift
• Characterizing Adversarial Subspaces Using Local Intrinsic Dimensionality
• Deep Anomaly Detection with Outlier Exposure
• Detecting Adversarial Samples from Artifacts
• Learning Confidence for Out-of-Distribution Detection in Neural Networks
• Importance Weighted Autoencoders
• WAIC, but Why? Generative Ensembles for Robust Anomaly Detection
• Constructing Unrestricted Adversarial Examples with Generative Models
• Automated interpretation of congenital heart disease from multi-view echocardiograms
• Quality of Uncertainty Quantification for Bayesian Neural Network Inference
• Detecting Out-of-Distribution Inputs to Deep Generative Models Using Typicality
• AUTO3D: Novel view synthesis through unsupervisely learned variational viewpoint and global 3D representation
• Conservative Wasserstein Training for Pose Estimation
• Disentanglement for Discriminative Visual Recognition

Cited by in corpus (14)

• Automated interpretation of congenital heart disease from multi-view echocardiograms
• Run-Time Monitoring of Machine Learning for Robotic Perception: A Survey of Emerging Trends
• Understanding Anomaly Detection with Deep Invertible Networks through Hierarchies of Distributions and Features
• AUTO3D: Novel view synthesis through unsupervisely learned variational viewpoint and global 3D representation
• Conservative Wasserstein Training for Pose Estimation
• Calibrating Deep Neural Network Classifiers on Out-of-Distribution Datasets
• Disentanglement for Discriminative Visual Recognition
• An evaluation of word-level confidence estimation for end-to-end automatic speech recognition
• Increasing Trustworthiness of Deep Neural Networks via Accuracy Monitoring
• Energy-constrained Self-training for Unsupervised Domain Adaptation
• Identity-aware Facial Expression Recognition in Compressed Video
• Self-Supervised Dynamic Networks for Covariate Shift Robustness
• Importance-Aware Semantic Segmentation in Self-Driving with Discrete Wasserstein Training
• DOI: Divergence-based Out-of-Distribution Indicators via Deep Generative Models