Agnostic Active Learning Without Constraints

Abstract

We present and analyze an agnostic active learning algorithm that works without keeping a version space. This is unlike all previous approaches where a restricted set of candidate hypotheses is maintained throughout learning, and only hypotheses from this set are ever returned. By avoiding this version space approach, our algorithm sheds the computational burden and brittleness associated with maintaining version spaces, yet still allows for substantial improvements over supervised learning for classification.

Cited by in corpus (39)

• Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds
• A Benchmark and Comparison of Active Learning for Logistic Regression
• Making Contextual Decisions with Low Technical Debt
• Active Learning from Weak and Strong Labelers
• Online Importance Weight Aware Updates
• Activized Learning: Transforming Passive to Active with Improved Label Complexity
• Active Learning from Imperfect Labelers
• Active Learning for Crowd-Sourced Databases
• A Unified View of Regularized Dual Averaging and Mirror Descent with Implicit Updates
• Exploring Connections Between Active Learning and Model Extraction
• Noisy Bayesian Active Learning
• Efficient and Parsimonious Agnostic Active Learning
• Asymptotic Analysis of Objectives based on Fisher Information in Active Learning
• Consistency-based Semi-supervised Active Learning: Towards Minimizing Labeling Cost
• Low-Cost Learning via Active Data Procurement
• AUTOLYCUS: Exploiting Explainable AI (XAI) for Model Extraction Attacks against Interpretable Models
• Active Learning Using Smooth Relative Regret Approximations with Applications
• Limitations of Assessing Active Learning Performance at Runtime
• Active Learning Algorithms for Graphical Model Selection
• Online Active Model Selection for Pre-trained Classifiers
• Active Clustering with Model-Based Uncertainty Reduction
• Search Improves Label for Active Learning
• Exploiting Context for Robustness to Label Noise in Active Learning
• Para-active learning
• A Compression Technique for Analyzing Disagreement-Based Active Learning
• Adaptive Region-Based Active Learning
• Active Learning under Label Shift
• Improved Algorithms for Agnostic Pool-based Active Classification
• The Utility of Abstaining in Binary Classification
• On the Power of Localized Perceptron for Label-Optimal Learning of Halfspaces with Adversarial Noise
• Flattening a Hierarchical Clustering through Active Learning
• Analyzing Data Selection Techniques with Tools from the Theory of Information Losses
• Active Online Learning with Hidden Shifting Domains
• Improved Algorithms for Efficient Active Learning Halfspaces with Massart and Tsybakov noise
• Corruption Robust Active Learning
• Distribution Matching for Machine Teaching
• Skyblocking for Entity Resolution
• Beating the Minimax Rate of Active Learning with Prior Knowledge
• Distribution-Dependent Sample Complexity of Large Margin Learning