Open Internet by MindsNet

Relevance of Active Learning in the Era of Large Models

The author questions whether active learning remains relevant or has become outdated in the context of recent advancements in large models and foundation models. Despite its potential to reduce labeled samples and annotation costs, active learning seems to be overlooked in industry and academia, with many researchers still relying on traditional methods like ResNet-18 for image classification tasks.

Computing & Technology, Computer Science, Machine Learning