Six Sigma for Neural Networks: Taguchi-Based Optimization

Sai Varun Kodathala · 2025

Read Six Sigma for Neural Networks: Taguchi-Based Optimization by Sai Varun Kodathala (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-2509-25213.txt.

Description

Sai Varun Kodathala applies quality engineering methods to neural networks in a 2025 study. Tuning a network's settings by trial and error is slow and costly. The author instead uses the Taguchi method, a design from factory quality work that tests a small, smart set of setting mixes.

The test task was spotting actions in professional boxing videos. Eight settings, from image size to learning rate, were varied across twelve runs. Five scoring schemes balanced training and test accuracy against error rates. The best scheme reached high training accuracy and solid test accuracy with low loss. Learning rate mattered most, then image size and the choice of activation function, giving clear guidance on what to tune first.

Catalog record

CREATORSai Varun Kodathala
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-25213

Browse the full library · About the Tower of Babel