Active Learning with Statistical Models

Abstract

For many types of machine learning algorithms, one can compute the statistically `optimal' way to select training data. In this paper, we review how optimal data selection techniques have been used with feedforward neural networks. We then show how the same principles may be used to select data for two alternative, statistically-based learning architectures: mixtures of Gaussians and locally weighted regression. While the techniques for neural networks are computationally expensive and approximate, the techniques for mixtures of Gaussians and locally weighted regression are both efficient and accurate. Empirically, we observe that the optimality criterion sharply decreases the number of training examples the learner needs in order to achieve good performance.

See http://www.jair.org/ for any accompanying files

Cited by in corpus (24)

• Machine Learning of Linear Differential Equations using Gaussian Processes
• A survey of active learning algorithms for supervised remote sensing image classification
• Inferring solutions of differential equations using noisy multi-fidelity data
• Active Learning with Multiple Views
• Committee-Based Sample Selection for Probabilistic Classifiers
• Long-Term Personalization of an In-Home Socially Assistive Robot for Children With Autism Spectrum Disorders
• Active Learning in Robotics: A Review of Control Principles
• O-MedAL: Online Active Deep Learning for Medical Image Analysis
• The Data Representativeness Criterion: Predicting the Performance of Supervised Classification Based on Data Set Similarity
• Active Learning of Discriminative Subgraph Patterns for API Misuse Detection
• On the Entropy Computation of Large Complex Gaussian Mixture Distributions
• Interpolating Detailed Simulations of Kilonovae: Adaptive Learning and Parameter Inference Applications
• Multimodal Hierarchical Dirichlet Process-based Active Perception
• Model-Centric and Data-Centric Aspects of Active Learning for Deep Neural Networks
• Probabilistic Inference from Arbitrary Uncertainty using Mixtures of Factorized Generalized Gaussians
• Active Learning of Abstract Plan Feasibility
• Combining Federated and Active Learning for Communication-efficient Distributed Failure Prediction in Aeronautics
• Data-adaptive Active Sampling for Efficient Graph-Cognizant Classification
• Learning and Extrapolation of Robotic Skills using Task-Parameterized Equation Learner Networks
• Impact of Feedback Type on Explanatory Interactive Learning
• Online A-Optimal Design and Active Linear Regression
• Cyber Situation Awareness with Active Learning for Intrusion Detection
• A unified framework for manifold landmarking
• Problem-fluent models for complex decision-making in autonomous materials research