An Information Theory Approach on Deciding Spectroscopic Follow-ups

Abstract

Classification and characterization of variable phenomena and transient phenomena are critical for astrophysics and cosmology. These objects are commonly studied using photometric time series or spectroscopic data. Given that many ongoing and future surveys are in time-domain and given that adding spectra provide further insights but requires more observational resources, it would be valuable to know which objects should we prioritize to have spectrum in addition to time series. We propose a methodology in a probabilistic setting that determines a-priory which objects are worth taking spectrum to obtain better insights, where we focus insight as the type of the object (classification). Objects for which we query its spectrum are reclassified using their full spectrum information. We first train two classifiers, one that uses photometric data and another that uses photometric and spectroscopic data together. Then for each photometric object we estimate the probability of each possible spectrum outcome. We combine these models in various probabilistic frameworks (strategies) which are used to guide the selection of follow up observations. The best strategy depends on the intended use, whether it is getting more confidence or accuracy. For a given number of candidate objects (127, equal to 5% of the dataset) for taking spectra, we improve 37% class prediction accuracy as opposed to 20% of a non-naive (non-random) best base-line strategy. Our approach provides a general framework for follow-up strategies and can be extended beyond classification and to include other forms of follow-ups beyond spectroscopy.

A Reinforcement Learning–Based Follow-up Framework

“…In our setting, the P model is fitted and tested over an independent portion of the data set before carrying on our RL framework for action recommendations (the data split details are described in Section 7.1). Furthermore, P corresponds to a collection of random forest classifiers (Breiman et al 1984), with a dedicated classifier for each possible data collection scenario, following the same pattern as in our previous work (Astudillo et al 2020). We chose to use random forest as our base classifier based on its robustness, fast training, and consistently good performance in various fields, including astronomical object classification.…”

“…Synoptic sky surveys, which repeatedly observe vast sky areas, have transformed the way the astrophysics community conducts research and discoveries. The enormous number of objects and the uniformity of the data enable systematic studies and feed the candidate pool for more costly follow-up studies of interesting objects (Djorgovski et al 2013;Astudillo et al 2020), such as variable or transient phenomena.…”

“…It is closely related to our primary goal: optimizing observational decisions in near-real time. Previously, we designed a probabilistic setting framework to target spectroscopic followups given a photometric time series (Astudillo et al 2020). This framework was mainly concerned with selecting the objects worth taking a spectrum observation of on top of their photometric ones to obtain better insights, focusing on a classification sense.…”