Active Learning

Active learning aims to improve learning efficiency by allowing the learning algorithm to select which data to learn from.

How it works

Traditional supervised learning often requires a large number of labeled instances, which can be costly or time-consuming to obtain. Active learning addresses this labeling bottleneck by asking an oracle (e.g., a human annotator) to label selected unlabeled instances. The goal is to achieve high accuracy with minimal labeling effort.

Considerations

Active learning is particularly useful in scenarios where data is abundant but labeled instances are scarce or expensive. Examples include speech recognition, information extraction, and document classification.

References

• Wikipedia article on Active Learning (machine learning) Link
• Settles, B. (2009). Active Learning Literature Survey. Link