Deep Transfer Learning for Classification of Variable Sources

TL;DR

This paper develops a deep neural network classifier for variable star light-curves and demonstrates that transfer learning allows effective adaptation to new datasets with minimal labeled data, improving classification across astronomical surveys.

Contribution

The paper introduces a transfer learning approach for deep neural network classifiers, enabling effective adaptation to new astronomical datasets with limited labeled data.

Findings

Transfer learning improves classification accuracy on new datasets.

Knowledge transfer reduces the need for extensive labeled data.

Method works across multiple astronomical surveys.

Abstract

Ongoing or upcoming surveys such as Gaia, ZTF, or LSST will observe light-curves of billons or more astronomical sources. This presents new challenges for identifying interesting and important types of variability. Collecting a sufficient number of labelled data for training is difficult, however, especially in the early stages of a new survey. Here we develop a single-band light-curve classifier based on deep neural networks, and use transfer learning to address the training data paucity problem by conveying knowledge from one dataset to another. First we train a neural network on 16 variability features extracted from the light-curves of OGLE and EROS-2 variables. We then optimize this model using a small set (e.g. 5%) of periodic variable light-curves from the ASAS dataset in order to transfer knowledge inferred from OGLE/EROS-2 to a new ASAS classifier. With this we achieve good classification results on ASAS, thereby showing that knowledge can be successfully transferred between datasets. We demonstrate similar transfer learning using Hipparcos and ASAS-SN data. We therefore find that it is not necessary to train a neural network from scratch for every new survey, but rather that transfer learning can be used even when only a small set of labelled data is available in the new survey.