Active Learning Strategies for Data-efficient Training: Investigating active learning strategies to select the most informative data samples for model training

Abstract

Active learning is a machine learning paradigm that aims to reduce the amount of labeled data required for model training by selecting the most informative data samples for annotation. This paper provides a comprehensive overview of active learning strategies for data-efficient training. We discuss various active learning approaches, including uncertainty sampling, query by committee, and diversity-based sampling. Additionally, we explore the application of active learning in different domains, such as image classification, text classification, and object detection. Through an extensive literature review, we highlight the effectiveness of active learning in improving model performance with limited labeled data. We also discuss challenges and future research directions in active learning for data-efficient training.