A kinetic-based regularization method for data science applications

Abhisek Ganguly, Alessandro Gabbana, Vybhav Rao, Sauro Succi, Santosh Ansumali · 2025

Read A kinetic-based regularization method for data science applications by Abhisek Ganguly, Alessandro Gabbana, Vybhav Rao, Sauro Succi, Santosh Ansumali (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-04857.txt.

Description

Abhisek Ganguly, Alessandro Gabbana, Vybhav Rao, Sauro Succi, and Santosh Ansumali borrow from statistical mechanics to keep function learning honest. Fitting an interpolator resembles minimizing a physical system's energy. Their regularization adds corrections that pin down the low-order moments of the data distribution. Those constraints shrink the gap between the discrete data and its continuum shadow.

Cleaner moment matching opens friendlier energy landscapes for the optimizer to descend. Accuracy improves on both interpolation and regression tasks, even in high-dimensional spaces. No empirical tuning knobs are needed, a relief with noisy data. The method's local nature also beats radial basis function interpolators on compute and memory for large data sets. Physics here is not decoration: the moment constraints do measurable work.

Catalog record

CREATORAbhisek Ganguly, Alessandro Gabbana, Vybhav Rao, Sauro Succi, Santosh Ansumali
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-04857

Browse the full library · About the Tower of Babel