A neural network-based methodology to select young stellar object candidates from IR surveys

TL;DR

This paper presents a machine learning approach using neural networks to classify young stellar objects from infrared survey data, achieving high accuracy and providing a reproducible methodology for astrophysical classification.

Contribution

The study introduces a neural network-based method for YSO classification from IR data, demonstrating high accuracy and generalization across star-forming regions, with publicly available catalogs.

Findings

Achieved over 90% recovery rate for CI YSOs

Achieved over 97% recovery rate for CII YSOs

Provided publicly accessible YSO candidate catalogs

Abstract

Observed Young Stellar Objects (YSOs) are used to study star formation and characterize star forming regions. For this purpose, YSO candidate catalogs are compiled from various surveys, especially in the infrared (IR), and simple selection schemes in colour-magnitude diagrams (CMDs) are often used to identify and classify YSOs. We propose a methodology for YSO classification through Machine Learning (ML) using Spitzer IR data. We detail our approach in order to ensure reproducibility and provide an in-depth example on how to efficiently apply ML to an astrophysical classification. We used feedforward Artificial Neural Networks (ANNs) that use the four IRAC bands ($3.6, 4.5, 5.8$ and $8 \mu m$) and the $24\ \mu m$ MIPS band from Spitzer to classify point source objects into CI and CII YSO candidates or as contaminants. We found that ANNs can efficiently be applied to YSO classification with a contained number of neurons ($\sim$ 25). Knowledge gathered on one star-forming region has shown to be partly efficient for prediction in new regions. Best generalization capacity was achieved using a combination of several star-forming regions to train the network. Carefully rebalancing the training proportions was necessary to achieve good results. We achieved above 90% and 97% recovery rate for CI and CII YSOs, respectively, with precision above 80% and 90% for our most general result. We took advantage of the ANN great flexibility to define, for each object, an effective membership probability to each output class. Using a threshold in this probability was found to efficiently improve the classification results at a reasonable cost of object exclusion. With this selection, we reached 90% precision on CI YSOs, for more than half of them. Our catalog of YSO candidates in Orion (365 CI, 2381 CII) and NGC 2264 (101 CI, 469 CII) predicted by our final ANN is publicly available at CDS.