Oral Abstract

Oral Contribution (O4.4) Shraddha Surana (ThoughtWorks Pvt. Ltd.)

Theme: Data science challenges: tools from statistics to machine learning

Machine Learning for Scientific Discovery

Machine learning algorithms are a very good tool for both classification and prediction purposes. However, these algorithms can be further used for scientific discovery from the enormous data being collected in our era. We present ways of discovering physical phenomenon by using machine learning algorithms on data collected through radio telescopes. We will discuss the use of supervised machine learning algorithms to predict the free parameters of star formation histories and also better understand the relations between the different input and output parameters. We made use of Deep Learning to capture the non-linearity in the parameters. Our models are able to predict with low error rates and gives the advantage of predicting in real time once the model has been trained.
The other class of machine learning algorithms viz. unsupervised learning prove to be very useful to uncover patterns in the data. We explore how we used such unsupervised techniques on solar radio telescope data to identify patterns and anomalies and also link such findings to theories, which help to better understand the nature of the elements being studies. We highlight the challenges faced in terms of data size, availability, features, processing ability and importantly, the interpretability of results. As our ability to capture & store data increases, it is inevitable, to use machine learning to understand the underlying information captured.