Radio Galaxy Zoo: Using semi-supervised learning to leverage large unlabelled data-sets for radio galaxy classification under data-set shift

TL;DR

This study evaluates semi-supervised learning for radio galaxy classification, revealing its benefits, limitations under dataset shift, and the challenges in performance prediction using data-set shift measures.

Contribution

It demonstrates the performance and calibration limitations of SSL in radio galaxy classification, especially under dataset shift and class imbalance, and explores data-set shift measurement techniques.

Findings

SSL outperforms baseline accuracy within a narrow label volume range

SSL does not improve model calibration regardless of accuracy gains

Data-set shift significantly reduces SSL performance when training and unlabeled data differ

Abstract

In this work we examine the classification accuracy and robustness of a state-of-the-art semi-supervised learning (SSL) algorithm applied to the morphological classification of radio galaxies. We test if SSL with fewer labels can achieve test accuracies comparable to the supervised state-of-the-art and whether this holds when incorporating previously unseen data. We find that for the radio galaxy classification problem considered, SSL provides additional regularisation and outperforms the baseline test accuracy. However, in contrast to model performance metrics reported on computer science benchmarking data-sets, we find that improvement is limited to a narrow range of label volumes, with performance falling off rapidly at low label volumes. Additionally, we show that SSL does not improve model calibration, regardless of whether classification is improved. Moreover, we find that when different underlying catalogues drawn from the same radio survey are used to provide the labelled and unlabelled data-sets required for SSL, a significant drop in classification performance is observered, highlighting the difficulty of applying SSL techniques under dataset shift. We show that a class-imbalanced unlabelled data pool negatively affects performance through prior probability shift, which we suggest may explain this performance drop, and that using the Frechet Distance between labelled and unlabelled data-sets as a measure of data-set shift can provide a prediction of model performance, but that for typical radio galaxy data-sets with labelled sample volumes of O(1000), the sample variance associated with this technique is high and the technique is in general not sufficiently robust to replace a train-test cycle.