CAASTRO student's DASH will speed up supernova classification
8 December 2016
CAASTRO student Daniel Muthukrishna has improved the speed at which supernova can be classified by building a program that works in the same way as the human brain – and has won an award for this project.
Daniel, a software engineering student at the University of Queensland, has received the 2016 Student Thesis prize given by the Queensland chapter of the Institute of Electrical and Electronics Engineers (IEEE).
For his thesis, Daniel developed new software for a CAASTRO-supported project, the OzDES redshift survey. Called DASH, the software greatly speeds up the process of classifying supernova taken from OzDES observations.
The code uses Deep Learning, a machine learning technique where the program is trained to look for patterns in the supernova templates and use these to identify the new spectra, but without these patterns being identified beforehand. The machine can learn for itself, in the same manner as a human brain can.
The code has already been used to analyse spectra taken by the AAT and returns the same results as the standard human-intensive process, but in a fraction of the time. Although the code has been developed to help with the OzDES analysis, it is very generic in the analysis approach, and would be usable by any spectroscopic supernova survey.
Daniel's thesis also won the GBST Prize for "Best software project prize" at the UQ Innovation Showcase event in November. He is currently on a ten-week research placement at the Gemini South Observatory in Chile, as one of the two Australian Gemini Undergraduate Summer Students for 2016.