Skip to main content
SearchLoginLogin or Signup

Exploring Charge Exchange Emission in NGC253 Through Automatic Spectroscopic Fitting using Neuroevolution algorithms

Presentation #104.17 in the session ISM/Galaxies - Poster Session.

Published onMay 03, 2024
Exploring Charge Exchange Emission in NGC253 Through Automatic Spectroscopic Fitting using Neuroevolution algorithms

We developed an automated spectroscopic analysis tool using Neuroevolution algorithms (NEA) to search the global optimum of fitted spectra. It is a process involving reduced human intervention but improved efficiency and reproducibility. This is achieved by taking advantage of both the evolution algorithm (EA) and convolutional neural networks (CNNs) and is vital in the era when current and upcoming detectors acquire massive data at rates orders of magnitude greater than current collection rates. The EA is a set of metaheuristic methods for evaluating the optimization problem of fitting and obtaining a global optimum of parameters from various physics models. We selected deferential evolution (DE) and genetic algorithms (GA) due to their performance. We start with a population of temporary solutions consisting of model parameters called chromosomes. We then select part of the best solutions by evaluating their fitness value and random solutions for mixing, i.e., crossover for the next generation of solutions. Next, we apply mutation operators to modify existing solutions by disturbing them by random chance. We also provide several types of selection operators to avoid ill-condition the EA operators may encounter in specific applications. We then switch to use EA-trained NN to continue searches in a more efficient way. The method has been applied to X-ray spectral data of a starburst galaxy NGC 253 from XMM-Newton reflection grating spectrometer, and compared with direct fitting using Xspec and Sherpa. The fitting considered a complicated physical model that includes thermal emission from the diffuse hot plasma as well as the charge exchange emission due to its interaction with the cold gas, besides the normal components of the bright point sources and the foreground absorption. The results have been compared with both MCMC methods and our early studies using only GA and show advantages in performance and accuracy.

Comments
0
comment
No comments here