Going beyond simple error analysis of ML systems

A few months ago I wrote an article about some considerations of how to perform error analysis for a trained machine learning model. It also proposes a “framework”/progression ladder for what methods to implement first.

I’d be happy if some of you find it useful, or maybe has some comments/questions.

analysis error machinelearningnews systems