Chinmay S. Kulkarni | Searcharxiv

arXiv2026

27 years of Spaceborne IR Astronomy: An ISO, Spitzer, WISE and NEOWISE Survey for Large-Amplitude Variability in Young Stellar Objects

Infrared observations can probe photometric variability across the full evolutionary range of young stellar objects (YSOs), from deeply embedded protostars to pre-main-sequence stars with dusty disks. We present 3-8 micron light curves extending 27 years from 1997 to 2024 obtained with three space-based IR telescopes: ISO, Spitzer and WISE. Although unevenly sampled with large gaps in coverage, these light curves show variability on time scales ranging from days to decades. We focus on the Spitzer-identified YSOs with disks and envelopes that exhibit variations of a factor of two or more in this wavelength range. We identified seven YSOs where the light curves are dominated by bursts of sustained (> 5 yr) high flux, including four that show a steep decay ending the burst and three that are ongoing as of the final observation. We find six YSOs that are undergoing declines, which may be the end of bursts that began before 1997. The most common form of variability, exhibited by 26 YSOs in our sample, show variations over time intervals of years to months but do not exhibit sustained bursts or fades. The Spitzer [3.6]-[4.5] and WISE [3.5]-[4.6] colors either increase or remain constant with increasing brightness, inconsistent with dust extinction as being the primary source of the large-amplitude variability.

astro-ph.SR↗

arXiv2019

Sparse Regression and Adaptive Feature Generation for the Discovery of Dynamical Systems

We study the performance of sparse regression methods and propose new techniques to distill the governing equations of dynamical systems from data. We first look at the generic methodology of learning interpretable equation forms from data, proposed by Brunton et al., followed by performance of LASSO for this purpose. We then propose a new algorithm that uses the dual of LASSO optimization for higher accuracy and stability. In the second part, we propose a novel algorithm that learns the candidate function library in a completely data-driven manner to distill the governing equations of the dynamical system. This is achieved via sequentially thresholded ridge regression (STRidge) over a orthogonal polynomial space. The performance of the three discussed methods is illustrated by looking the Lorenz 63 system and the quadratic Lorenz system.

cs.LG↗