Bayesian Active Learning for Classification and Preference Learning

Abstract

Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classification with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classifier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classification problem, we extend our algorithm to Gaussian Process preference learning.

Cited by in corpus (138)

• A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges
• Deep Bayesian Active Learning with Image Data
• BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning
• Uncertainty Estimation Using a Single Deep Deterministic Neural Network
• Galaxy Zoo: Probabilistic Morphology through Bayesian CNNs and Active Learning
• Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples
• Controversial stimuli: pitting neural networks against each other as models of human recognition
• Active label cleaning for improved dataset quality under resource constraints
• Selection via Proxy: Efficient Data Selection for Deep Learning
• Bayesian Batch Active Learning as Sparse Subset Approximation
• A Survey of Active Learning for Text Classification using Deep Neural Networks
• Domain-independent Extraction of Scientific Concepts from Research Articles
• Asking Easy Questions: A User-Friendly Approach to Active Reward Learning
• A Unifying Framework for Gaussian Process Pseudo-Point Approximations using Power Expectation Propagation
• A Bayesian Framework for Digital Twin-Based Control, Monitoring, and Data Collection in Wireless Systems
• Uncertainty-aware Self-training for Text Classification with Few Labels
• Transient-optimised real-bogus classification with Bayesian Convolutional Neural Networks -- sifting the GOTO candidate stream
• Identifying Causal-Effect Inference Failure with Uncertainty-Aware Models
• On Statistical Bias In Active Learning: How and When To Fix It
• Information Aware Max-Norm Dirichlet Networks for Predictive Uncertainty Estimation
• Adversarial Distillation of Bayesian Neural Network Posteriors
• Active Learning of Linear Embeddings for Gaussian Processes
• Uncertainty as a Form of Transparency: Measuring, Communicating, and Using Uncertainty
• Adversarial Phenomenon in the Eyes of Bayesian Deep Learning
• Deep Active Learning for Computer Vision: Past and Future
• Deep Ensemble Bayesian Active Learning : Addressing the Mode Collapse issue in Monte Carlo dropout via Ensembles
• Reinforced active learning for image segmentation
• Uncertainty Based Detection and Relabeling of Noisy Image Labels
• Towards Active Learning for Action Spotting in Association Football Videos
• Inhibited Softmax for Uncertainty Estimation in Neural Networks
• Local Augmentation for Graph Neural Networks
• Task-Aware Variational Adversarial Active Learning
• Adversarial Representation Active Learning
• Bayesian Active Meta-Learning for Few Pilot Demodulation and Equalization
• Prior and Posterior Networks: A Survey on Evidential Deep Learning Methods For Uncertainty Estimation
• Active Learning for Visual Question Answering: An Empirical Study
• Scaling Hamiltonian Monte Carlo Inference for Bayesian Neural Networks with Symmetric Splitting
• Model-Centric and Data-Centric Aspects of Active Learning for Deep Neural Networks
• A Weakly Supervised Region-Based Active Learning Method for COVID-19 Segmentation in CT Images
• Active Learning for Deep Detection Neural Networks
• Bayesian Variational Autoencoders for Unsupervised Out-of-Distribution Detection
• Self-Supervised Exploration via Disagreement
• Active Testing: Sample-Efficient Model Evaluation
• Adaptive Nonparametric Psychophysics
• Actively Learning what makes a Discrete Sequence Valid
• Small-Text: Active Learning for Text Classification in Python
• Active Learning of Abstract Plan Feasibility
• Exploring Spatial Diversity for Region-based Active Learning
• Deep Active Learning for Text Classification with Diverse Interpretations
• Deep Deterministic Uncertainty for Semantic Segmentation
• ActiveHARNet: Towards On-Device Deep Bayesian Active Learning for Human Activity Recognition
• Deep Bayesian Active Learning for Multiple Correct Outputs
• On uncertainty estimation in active learning for image segmentation
• Active Learning Solution on Distributed Edge Computing
• Dense Uncertainty Estimation
• Bayesian Estimate of Mean Proper Scores for Diversity-Enhanced Active Learning
• Evidential Deep Learning for Open Set Action Recognition
• Bayesian active learning for production, a systematic study and a reusable library
• Bayesian Active Learning for Discrete Latent Variable Models
• Learning Human Objectives by Evaluating Hypothetical Behavior
• Icebreaker: Element-wise Active Information Acquisition with Bayesian Deep Latent Gaussian Model
• Unpacking Information Bottlenecks: Unifying Information-Theoretic Objectives in Deep Learning
• Reducing Annotating Load: Active Learning with Synthetic Images in Surgical Instrument Segmentation
• Stopping Criterion for Active Learning Based on Error Stability
• GeneDisco: A Benchmark for Experimental Design in Drug Discovery
• Efficient Exploration in Binary and Preferential Bayesian Optimization
• Composite Bayesian Optimization In Function Spaces Using NEON -- Neural Epistemic Operator Networks
• Causal-BALD: Deep Bayesian Active Learning of Outcomes to Infer Treatment-Effects from Observational Data
• Model-Change Active Learning in Graph-Based Semi-Supervised Learning
• Deep Bayesian Active Learning, A Brief Survey on Recent Advances
• Uncertainty aware audiovisual activity recognition using deep Bayesian variational inference
• Bayesian Active Learning for Wearable Stress and Affect Detection
• Sequential Graph Convolutional Network for Active Learning
• Forgetful Active Learning with Switch Events: Efficient Sampling for Out-of-Distribution Data
• Deep Multi-Fidelity Active Learning of High-dimensional Outputs
• Test Distribution-Aware Active Learning: A Principled Approach Against Distribution Shift and Outliers
• Active Preference-Based Gaussian Process Regression for Reward Learning
• Active learning for interactive satellite image change detection
• Deep Probabilistic Models to Detect Data Poisoning Attacks
• AIDA: An Active Inference-based Design Agent for Audio Processing Algorithms
• Noisy Batch Active Learning with Deterministic Annealing
• BAR: Bayesian Activity Recognition using variational inference
• Can Active Learning Preemptively Mitigate Fairness Issues?
• Single-Modal Entropy based Active Learning for Visual Question Answering
• Active${}^2$ Learning: Actively reducing redundancies in Active Learning methods for Sequence Tagging and Machine Translation
• Automated and Autonomous Experiment in Electron and Scanning Probe Microscopy
• Knowing what to know: Implications of the choice of prior distribution on the behavior of adaptive design optimization
• Learning to Map for Active Semantic Goal Navigation
• The Advantage of Cross Entropy over Entropy in Iterative Information Gathering
• Textual Membership Queries
• Hierarchical Gaussian Process Priors for Bayesian Neural Network Weights
• Experiment Design with Gaussian Process Regression with Applications to Chance-Constrained Control
• Active Deep Learning on Entity Resolution by Risk Sampling
• A Confidence-based Acquisition Model for Self-supervised Active Learning and Label Correction
• Effective Evaluation of Deep Active Learning on Image Classification Tasks
• A Bayesian binary classification approach to pure tone audiometry
• Super-resolution of Time-series Labels for Bootstrapped Event Detection
• One-bit Supervision for Image Classification: Problem, Solution, and Beyond
• Bayes-Optimal Entropy Pursuit for Active Choice-Based Preference Learning
• CLOPS: Continual Learning of Physiological Signals
• Information Condensing Active Learning
• Offline Preference-Based Apprenticeship Learning
• Bayesian Active Learning by Disagreements: A Geometric Perspective
• Efficient Uncertainty Estimation for Semantic Segmentation in Videos
• Active Learning in CNNs via Expected Improvement Maximization
• Semi-supervised Learning of Galaxy Morphology using Equivariant Transformer Variational Autoencoders
• Looking at the posterior: accuracy and uncertainty of neural-network predictions
• Contextual Bayesian optimization with binary outputs
• Robust recognition and exploratory analysis of crystal structures via Bayesian deep learning
• PowerEvaluationBALD: Efficient Evaluation-Oriented Deep (Bayesian) Active Learning with Stochastic Acquisition Functions
• LADA: Look-Ahead Data Acquisition via Augmentation for Active Learning
• High Dimensional Level Set Estimation with Bayesian Neural Network
• Towards Understanding the Behaviors of Optimal Deep Active Learning Algorithms
• Interactive Text Ranking with Bayesian Optimisation: A Case Study on Community QA and Summarisation
• Active Imitation Learning from Multiple Non-Deterministic Teachers: Formulation, Challenges, and Algorithms
• Effective Version Space Reduction for Convolutional Neural Networks
• NE-LP: Normalized Entropy and Loss Prediction based Sampling for Active Learning in Chinese Word Segmentation on EHRs
• Efficient Graph-Based Active Learning with Probit Likelihood via Gaussian Approximations
• Uncertainty quantification for White Matter Hyperintensity segmentation detects silent failures and improves automated Fazekas quantification
• ROIAL: Region of Interest Active Learning for Characterizing Exoskeleton Gait Preference Landscapes
• Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering
• Knowledge-driven Active Learning
• Targeted Active Learning for Bayesian Decision-Making
• Deep Active Learning by Model Interpretability
• Assisted Robust Reward Design
• MetAL: Active Semi-Supervised Learning on Graphs via Meta Learning
• Beyond Marginal Uncertainty: How Accurately can Bayesian Regression Models Estimate Posterior Predictive Correlations?
• Prediction stability as a criterion in active learning
• A Practical & Unified Notation for Information-Theoretic Quantities in ML
• Choice functions based multi-objective Bayesian optimisation
• HumBugDB: A Large-scale Acoustic Mosquito Dataset
• Residual Overfit Method of Exploration
• Feedback Coding for Active Learning
• Active learning for medical code assignment
• Dynamic Feature Acquisition with Arbitrary Conditional Flows
• Customs Fraud Detection in the Presence of Concept Drift
• SoQal: Selective Oracle Questioning in Active Learning
• Meta-Active Learning for Node Response Prediction in Graphs