Learning Algorithms for Active Learning

TL;DR

This paper presents a meta-learning approach to automatically learn active learning algorithms by jointly optimizing data representation, item selection, and prediction construction across related tasks, demonstrated on Omniglot and MovieLens datasets.

Contribution

It introduces a novel model that learns active learning strategies through meta-learning, combining multiple components for improved task-specific data selection.

Findings

Effective in synthetic and real-world settings

Improves data efficiency in active learning

Generalizes across different datasets

Abstract

We introduce a model that learns active learning algorithms via metalearning. For a distribution of related tasks, our model jointly learns: a data representation, an item selection heuristic, and a method for constructing prediction functions from labeled training sets. Our model uses the item selection heuristic to gather labeled training sets from which to construct prediction functions. Using the Omniglot and MovieLens datasets, we test our model in synthetic and practical settings.