Semi supervised learning improves active learning with fixed training cost

Should Pseudo-Labels Share Mini-Batches with Labeled Examples? Semi-Supervised Continual Active Learning with Limited Training Cost

Machine LearningArtificial IntelligenceSound

Summary

Active learning helps reduce the effort needed to label data by focusing on the most useful examples. The authors study continual active learning, where the model is trained with a limited number of training steps in each round, making training more efficient. They find a better way to combine pseudo-labeled data (unlabeled data given predicted labels) with labeled data during training, ensuring the training cost stays fixed while benefiting from extra supervision. Their method, called SSCAL, performs well across various data and settings, showing that mixing pseudo-labels within mini-batches improves results without extra training time.

What this means in practice

  • For machine learning engineers: Reduce annotation and training time when building models for audio data by better combining pseudo-labels with labeled examples in active learning cycles.
  • For speech recognition developers: Improve continual training of speech models with limited computational resources by applying SSCAL to mix pseudo-labels efficiently in mini-batch training.

Authors

Shiqi Zhang, Tuomas Virtanen

Abstract

Active learning (AL) reduces annotation by querying informative examples. Continual active learning (CAL) also cuts the training cost, the total number of gradient steps, by training one model across AL rounds with a fixed number of mini-batch steps per round. Pseudo-labels add supervision without annotation, but they increase training cost. In CAL, where every round trains on a fixed number of mini-batches, whether and how pseudo-labeled examples should share these mini-batches with labeled examples is therefore open. We provide the first formalization of this problem as two decisions after each query, namely (i) which unlabeled examples should enter training with a pseudo-label and (ii) what share of each mini-batch they should take. Our method, semi-supervised continual active learning (SSCAL), works with any acquisition function. It admits pseudo-labels based on distances from unlabeled examples to labeled neighbors, caps their share, and reweights the loss on labeled examples to keep their expected gradient contribution unchanged. On frozen embeddings, its pseudo-labeler updates incrementally without extra model forward passes. Across two audio datasets, two frozen encoders, six acquisition functions and ten seeds, SSCAL ranks first on average among four protocols with equal training steps, answering the question affirmatively.
View PDFOpen arXiv