Authors: S Sabouri, A Zolanvari, S Haghighi Venue: arXiv, 2026
Summary
A tutorial on using the PyCM library for detailed evaluation of multi-class classifiers. Through two case studies, it shows that the choice of metrics can change how a model’s performance looks, and argues that evaluating models across several dimensions is needed to catch small but meaningful differences that standard metrics miss.
Links
Related
PyCM — Multi-Class Confusion Matrix Library: the library the tutorial is built on. Sepand Haghighi: co-author.