ImitAL: Learning Active Learning Strategies from Synthetic Data

TL;DR

ImitAL introduces a novel, learned active learning query strategy using imitation learning trained on synthetic data, demonstrating broad applicability, improved performance, and efficiency across diverse datasets and strategies.

Contribution

It presents ImitAL, a new active learning method that encodes query selection as a learning-to-rank problem trained via imitation learning on synthetic data.

Findings

Outperforms 10 state-of-the-art strategies on 15 datasets

More computationally efficient, especially on large datasets

Demonstrates broad domain applicability

Abstract

One of the biggest challenges that complicates applied supervised machine learning is the need for huge amounts of labeled data. Active Learning (AL) is a well-known standard method for efficiently obtaining labeled data by first labeling the samples that contain the most information based on a query strategy. Although many methods for query strategies have been proposed in the past, no clear superior method that works well in general for all domains has been found yet. Additionally, many strategies are computationally expensive which further hinders the widespread use of AL for large-scale annotation projects. We, therefore, propose ImitAL, a novel query strategy, which encodes AL as a learning-to-rank problem. For training the underlying neural network we chose Imitation Learning. The required demonstrative expert experience for training is generated from purely synthetic data. To show the general and superior applicability of \ImitAL{}, we perform an extensive evaluation comparing our strategy on 15 different datasets, from a wide range of domains, with 10 different state-of-the-art query strategies. We also show that our approach is more runtime performant than most other strategies, especially on very large datasets.