Artificial Intelligence

    

A Short Empirical Note on Scaling Behavior in Small Neural Networks

Authors: Ritvik Chappidi, Aditya Jupally

Scaling laws describe how model performance improves with dataset size, model width, and compute. While such laws are well documented for large-scale language models, their behavior in small networks remains less understood. This paper presents a concise empirical study of loss scaling behavior in simple feedforward neural networks trained on synthetic regression tasks. Results show that even very small networks follow an approximate power-law relationship between dataset size and test loss, with a fitted exponent of about 0.076. These findings suggest that scaling regularities emerge even at small scales, implying that the underlying principles of efficiency and generalization extend beyond large-scale models.

Comments: 2 Pages.

Download: PDF

Submission history

[v1] 2025-10-28 13:54:18

Unique-IP document downloads: 180 times

Vixra.org is a pre-print repository rather than a journal. Articles hosted may not yet have been verified by peer-review and should be treated as preliminary. In particular, anything that appears to include financial or legal advice or proposed medical treatments should be treated with due caution. Vixra.org will not be responsible for any consequences of actions that result from any form of use of any documents on this website.

Add your own feedback and questions here:
You are equally welcome to be positive or negative about any paper but please be polite. If you are being critical you must mention at least one specific error, otherwise your comment will be deleted as unhelpful.

comments powered by Disqus