TY - JOUR

T1 - Light-curve classification with recurrent neural networks for GOTO

T2 - dealing with imbalanced data

AU - Burhanudin, U F

AU - Maund, Justyn

AU - Killestein, Thomas

AU - Ackley, K

AU - Dyer, Martin J

AU - Lyman, Joe

AU - Ulaczyk, K

AU - Cutter, R.

AU - Mong, Y-L

AU - Steeghs, D

AU - Galloway, D K

AU - Dhillon, Vik

AU - O’Brien, P

AU - Ramsay, G

AU - Noysena, K

AU - Kotak, R

AU - Breton, Rene P

AU - Nuttall, L

AU - Pallé, E

AU - Pollacco, D

AU - Thrane, E

AU - Awiphan, S

AU - Chote, P

AU - Chrimes, A

AU - Daw, E

AU - Duffy, Christopher

AU - Eyles-Ferris, RAJ

AU - Gompertz, Benjamin

AU - Heikkilä, T

AU - Irawati, P

AU - Kennedy, Mark R

AU - Levan, A

AU - Littlefair, S

AU - Makrygianni, L

AU - Mata-Sánchez, D

AU - Mattila, S

AU - McCormac, J

AU - Mkrtichian, D

AU - Mullaney, J

AU - Sawangwit, U

AU - Stanway, Elizabeth

AU - Starling, Rhaana

AU - Strøm, P

AU - Tooke, S

AU - Wiersema, K

PY - 2021/8/31

Y1 - 2021/8/31

N2 - 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 overrepresented 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, 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.

AB - 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 overrepresented 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, 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.

U2 - 10.1093/mnras/stab1545

DO - 10.1093/mnras/stab1545

M3 - Journal article

VL - 505

SP - 4345

EP - 4361

JO - Monthly Notices of the Royal Astronomical Society

JF - Monthly Notices of the Royal Astronomical Society

SN - 0035-8711

IS - 3

ER -