Light curve classification with recurrent neural networks for GOTO: dealing with imbalanced data

Abstract

The advent of wide-field sky surveys has led to the growth of transient and variable source discoveries. The data deluge produced by these surveys has necessitated the use of machine learning (ML) and deep learning (DL) algorithms to sift through the vast incoming data stream. A problem that arises in real-world applications of learning algorithms for classification is imbalanced data, where a class of objects within the data is underrepresented, leading to a bias for over-represented classes in the ML and DL classifiers. We present a recurrent neural network (RNN) classifier that takes in photometric time-series data and additional contextual information (such as distance to nearby galaxies and on-sky position) to produce real-time classification of objects observed by the Gravitational-wave Optical Transient Observer (GOTO), and use an algorithm-level approach for handling imbalance with a focal loss function. The classifier is able to achieve an Area Under the Curve (AUC) score of 0.972 when using all available photometric observations to classify variable stars, supernovae, and active galactic nuclei. The RNN architecture allows us to classify incomplete light curves, and measure how performance improves as more observations are included. We also investigate the role that contextual information plays in producing reliable object classification.

LSST Cadence Strategy Evaluations for AGN Time-series Data in Wide-Fast-Deep Field

“…Connections between nodes form a directed graph along a temporal sequence. Dékány & Grebel (2020), Fremling et al (2021, Burhanudin et al (2021) Recurrent AutoEncoder (RAE) Its encoder learns a representation (encoding) for a set of data, typically for dimensional reduction (2D time sequences to 1D latent variables), by training the network to extract the inherent features from input sequences, and ignore insignificant or noisy data. Then, the representation, or so-called 1D latent variables are fed into a decoder, which decodes the features and generate output sequences.…”

“…We implement a stochastic recurrent neural network (SRNN) to model quasar light curves and recover the DRW and DHO model parameters by Gaussian Process Regression (GPR). Recurrent neural networks (RNN) are a popular class of ML connectionist models for sequential modelling and have been used previously in astrophysics applications (e.g., Charnock & Moss 2017;Naul et al 2018;Hinners et al 2018;Muthukrishna et al 2019;Möller & de Boissière 2020 Escamilla-Rivera et al 2020;Burhanudin et al 2021;Lin & Wu 2021). However, as noted in Yin & Barucca (2021), one limitation of RNNs is that the hidden state transition function is entirely deterministic, which can limit the RNNs ability to model processes with high variability.…”