A lightweight hybrid deep learning approach for fashion mnist classification with explainable attention visualization

Hafeez Ahmad, Tahira Anwar Lashari, Saima Anwar Lashari, Ijaz Khan, Farzana Jabeen · 2026

Read A lightweight hybrid deep learning approach for fashion mnist classification with explainable attention visualization by Hafeez Ahmad, Tahira Anwar Lashari, Saima Anwar Lashari, Ijaz Khan, Farzana Jabeen (2026) 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/plos-pone-0351671.txt.

Description

Hafeez Ahmad and colleagues shrank the brain needed to sort fashion photos into their ten clothing classes, a standard test problem for image software. Top models for the task carry tens of millions of settings and need graphics processors, shutting out cheap devices. Their design pairs a small convolutional network with a lean form of self-attention, keeping under half a million settings in total.

The small model trained on an ordinary processor and sorted the Fashion-MNIST benchmark with accuracy a little above ninety-one percent. The team also added several ways to picture where the model looks when it decides, so users can audit its attention. For shops and apps running on phones or shop-floor hardware, the message is that respectable image sorting no longer requires a data center, and the reasoning can be shown, not just trusted.

Catalog record

CREATORHafeez Ahmad, Tahira Anwar Lashari, Saima Anwar Lashari, Ijaz Khan, Farzana Jabeen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE156073
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0351671

Browse the full library · About the Tower of Babel