MTAKD: multi-teacher agreement knowledge distillation for edge AI skin disease diagnosis

Andreas Winata, Nur Afny Catur Andryani, Alexander Agung Santoso Gunawan, Ford Lumban Gaol, Tokuro Matsuo · 2025

Read MTAKD: multi-teacher agreement knowledge distillation for edge AI skin disease diagnosis by Andreas Winata, Nur Afny Catur Andryani, Alexander Agung Santoso Gunawan, Ford Lumban Gaol, Tokuro Matsuo (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/pmc-12722719.txt.

Description

Andreas Winata and co-authors present a 2025 paper on skin disease diagnosis on edge devices. Remote areas lack skin specialists and stable internet. Edge AI moves the diagnosis onto the phone itself. The team builds a distillation framework that teaches a small model from several large teachers. Teacher agreement signals how reliable each lesson is. Inputs where teachers agree get more weight.

Three new algorithms distill answers, attention maps, and feature relations under that agreement rule. Mean accuracy hits high marks on one standard skin set and lower marks on a harder, skin-tone-diverse set. The gap between sets is itself informative. Clinics get a mobile tool that works offline. Honest scores on the diverse set warn against deploying on accuracy from one benchmark alone. The agreement idea can travel to other edge diagnosis tasks.

Catalog record

CREATORAndreas Winata, Nur Afny Catur Andryani, Alexander Agung Santoso Gunawan, Ford Lumban Gaol, Tokuro Matsuo
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC12722719

Browse the full library · About the Tower of Babel