Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userA recurrent neural network for classification of unevenly sampled variable stars
94
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
Astronomical surveys of celestial sources produce streams of noisy time series measuring flux versus time (light curves). Unlike in many other physical domains, however, large (and source-specific) temporal gaps in data arise naturally due to intranight cadence choices as well as diurnal and seasonal constraints. With nightly observations of millions of variable stars and transients from upcoming surveys, efficient and accurate discovery and classification techniques on noisy, irregularly sampled data must be employed with minimal human-in-the-loop involvement. Machine learning for inference tasks on such data traditionally requires the laborious hand-coding of domain-specific numerical summaries of raw data (features). Here we present a novel unsupervised autoencoding recurrent neural network (RNN) that makes explicit use of sampling times and known heteroskedastic noise properties. When trained on optical variable star catalogs, this network produces supervised classification models that rival other best-in-class approaches. We find that autoencoded features learned on one time-domain survey perform nearly as well when applied to another survey. These networks can continue to learn from new unlabeled observations and may be used in other unsupervised tasks such as forecasting and anomaly detection.
rate research
Read More
Common variable star classifiers are built only with the goal of producing the correct class labels, leaving much of the multi-task capability of deep neural networks unexplored. We present a periodic light curve classifier that combines a recurrent neural network autoencoder for unsupervised feature extraction and a dual-purpose estimation network for supervised classification and novelty detection. The estimation network optimizes a Gaussian mixture model in the reduced-dimension feature space, where each Gaussian component corresponds to a variable class. An estimation network with a basic structure of a single hidden layer attains a cross-validation classification accuracy of ~99%, on par with the conventional workhorses, random forest classifiers. With the addition of photometric features, the network is capable of detecting previously unseen types of variability with precision 0.90, recall 0.96, and an F1 score of 0.93. The simultaneous training of the autoencoder and estimation network is found to be mutually beneficial, resulting in faster autoencoder convergence, and superior classification and novelty detection performance. The estimation network also delivers adequate results even when optimized with pre-trained autoencoder features, suggesting that it can readily extend existing classifiers to provide added novelty detection capabilities.
We present a machine learning package for the classification of periodic variable stars. Our package is intended to be general: it can classify any single band optical light curve comprising at least a few tens of observations covering durations from weeks to years, with arbitrary time sampling. We use light curves of periodic variable stars taken from OGLE and EROS-2 to train the model. To make our classifier relatively survey-independent, it is trained on 16 features extracted from the light curves (e.g. period, skewness, Fourier amplitude ratio). The model classifies light curves into one of seven superclasses - Delta Scuti, RR Lyrae, Cepheid, Type II Cepheid, eclipsing binary, long-period variable, non-variable - as well as subclasses of these, such as ab, c, d, and e types for RR Lyraes. When trained to give only superclasses, our model achieves 0.98 for both recall and precision as measured on an independent validation dataset (on a scale of 0 to 1). When trained to give subclasses, it achieves 0.81 for both recall and precision. In order to assess classification performance of the subclass model, we applied it to the MACHO, LINEAR, and ASAS periodic variables, which gave recall/precision of 0.92/0.98, 0.89/0.96, and 0.84/0.88, respectively. We also applied the subclass model to Hipparcos periodic variable stars of many other variability types that do not exist in our training set, in order to examine how much those types degrade the classification performance of our target classes. In addition, we investigate how the performance varies with the number of data points and duration of observations. We find that recall and precision do not vary significantly if the number of data points is larger than 80 and the duration is more than a few weeks. The classifier software of the subclass model is available from the GitHub repository (https://goo.gl/xmFO6Q).
The need for the development of automatic tools to explore astronomical databases has been recognized since the inception of CCDs and modern computers. Astronomers already have developed solutions to tackle several science problems, such as automatic classification of stellar objects, outlier detection, and globular clusters identification, among others. New science problems emerge and it is critical to be able to re-use the models learned before, without rebuilding everything from the beginning when the science problem changes. In this paper, we propose a new meta-model that automatically integrates existing classification models of variable stars. The proposed meta-model incorporates existing models that are trained in a different context, answering different questions and using different representations of data. Conventional mixture of experts algorithms in machine learning literature can not be used since each expert (model) uses different inputs. We also consider computational complexity of the model by using the most expensive models only when it is necessary. We test our model with EROS-2 and MACHO datasets, and we show that we solve most of the classification challenges only by training a meta-model to learn how to integrate the previous experts.
We present a novel automated methodology to detect and classify periodic variable stars in a large database of photometric time series. The methods are based on multivariate Bayesian statistics and use a multi-stage approach. We applied our method to the ground-based data of the TrES Lyr1 field, which is also observed by the Kepler satellite, covering ~26000 stars. We found many eclipsing binaries as well as classical non-radial pulsators, such as slowly pulsating B stars, Gamma Doradus, Beta Cephei and Delta Scuti stars. Also a few classical radial pulsators were found.
With recent developments in imaging and computer technology the amount of available astronomical data has increased dramatically. Although most of these data sets are not dedicated to the study of variable stars much of it can, with the application of proper software tools, be recycled for the discovery of new variable stars. Fits Viewer and Data Retrieval System is a new software package that takes advantage of modern computer advances to search astronomical data for new variable stars. More than 200 new variable stars have been found in a data set taken with the Calvin College Rehoboth Robotic telescope using FVDRS. One particularly interesting example is a very fast subdwarf B with a 95 minute orbital period, the fastest currently known of the HW Vir type.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
