Read the original document by opening this link in a new tab.
Table of Contents
Abstract
1 Introduction
1.1 Deployment of Machine Learning
1.2 Process of Active Learning
2 Preliminaries on Active Learning
2.1 Instance Selection Criteria
2.2 Active Learning Scenarios
2.2.1 Membership Query Synthesis Active Learning
2.2.2 Pool-based Active Learning
2.2.3 Online Active Learning
3 Evaluation Strategies for Online Active Learning Algorithms
4 Real-world Applications and Challenges
5 Summary of Online Active Learning Methods
6 Future Research Directions
7 Conclusions and Key Contributions
Summary
Active learning is a paradigm in machine learning that aims to select the most informative data points to label from a data stream. It helps minimize the cost associated with collecting labeled observations and improve the performance of machine learning models. The survey discusses various active learning approaches, including uncertainty-based query strategies, expected error minimization, diversity-based approaches, and hybrid strategies. It covers active learning scenarios such as membership query synthesis, pool-based active learning, and online active learning. The document also highlights evaluation strategies, real-world applications, and future research directions in the field of active learning.