arXiv subjects

Varun Satish

Publications and source records attributed to Varun Satish.

3 recordsLinked to original sources

arXiv2026

Births are difficult to predict even with rich survey and full-population register data

Major life events have proven difficult to predict. Does this reflect limits of theory, data, and algorithms, or the large role of chance? We examine one outcome - having a child within three years - through a near-ideal setting for prediction: a data challenge where 147 researchers predicted births for Dutch residents aged 18-45, using survey data and full-population registers. Methods ranged from logistic regression to a large language model and transformers. Predictions were moderately accurate (best F1: register 0.59, survey 0.76); advanced models did not outperform classical ones; and the larger registers did not beat the survey. Simulating the stochastic biology of conception and pregnancy, we estimated a predictive ceiling (survey F1 ~ 0.86-0.94, register 0.88-0.96). Observed performance falls short of this ceiling, implicating imperfect data, methods, and unmodelled chance, while the ceiling itself shows that chance in reproduction alone sets a non-trivial limit on predicting individual lives.

cs.LG↗

arXiv2026

The AGORA High-resolution Galaxy Simulations Comparison Project. XI: Solving the Non-Spherical Morphology and Evolution of Dark Matter Halos with Haskap Pie

We introduce a halo solving and tracking procedure that intrinsically treats dark matter halos as non-spherical objects by leveraging the bound particle searching techniques used in Haskap Pie. The AGORA Collaboration's hydrodynamic simulation CosmoRun}project provides a useful laboratory to explore trends in dark matter halo morphology that are revealed by our new procedure in the context of any dispersions or similarities between the codes. We find that several morphological and shape measures were very responsive to high mass ratio mergers. The greatest difference in these measures between the simulation codes were related to timing discrepancies and the dynamical state of the halos prior to the mergers. Most other quantities were similar across codes, including several secular and redshift-dependent trends in various dynamical quantities that showed a departure from Virial Theorem (e.g., overdensity and halo mass). We find that halo spin and the ratio between the semi-major and the semi-minor axis peaked at 4>z>2 before declining at low redshift. Also, halo overdensity is both mass-dependent and redshift-dependent, diverging for low mass halos at low redshift. Our method contributes a new perspective on these trends that have not been fully replicated in other works due to our emphasis on fundamentally non-spherical halos and measures of morphology that correspondingly do not assume spherical symmetry.

astro-ph.GA↗

arXiv2025

Effect of molecular hydrogen self-shielding modeling on early Reionization Era galaxies in radiative hydrodynamic cosmological simulations

Accurately modeling molecular hydrogen ($\text{H}_{2}$) is an important task in cosmological simulations because it regulates star formation. One fundamental property of $\text{H}_{2}$ is the ability to self-shield, a phenomenon in which the $\text{H}_{2}$ in the outer layer of a molecular cloud absorbs the photodissociating Lyman-Werner UV radiation and shields the inner $\text{H}_{2}$. Historically, numerical approximations have been utilized to avoid intensive ray-tracing calculations. This paper evaluates the use of the Sobolev-like density-gradient approximation in $\text{H}_{2}$ self-shielding modeling and tests its agreement with a more rigorous adaptive ray-tracing method in cosmological simulations. We ran four high-resolution zoom-in cosmological simulations to investigate the models' effects in the early Reionization Era ($z \geq 12$). We find that the approximation model returns a higher $\text{H}_{2}$ photodissociation rate in low gas density environments but a lower rate when gas density is high, resulting in low-mass halos having less $\text{H}_{2}$ while high-mass halos having more $\text{H}_{2}$. The approximation also hinders star formation in small halos, but it less affects the stellar mass of larger halos. Inside a halo, the discrepancies between the two models regarding $\text{H}_{2}$ fraction, temperature, and stellar mass are radially dependent. On a large scale, the simulations using the approximation have less $\text{H}_{2}$ in the intergalactic medium and may experience a slower reionization process. These results show that the Sobolev-like approximation alters properties of galaxies and the large-scale universe when compared to the ray-tracing treatment, emphasizing a need for caution when interpreting results from these two techniques in cosmological simulations.

astro-ph.GA↗