MinViME/Minimum Viable Model Estimator

Published in Software Impacts, 2021

Recommended citation: Hawkins, John. (2021) "MinViME/Minimum Viable Model Estimator" Software Impacts, Aug 01; Volume 9 https://doi.org/10.1016/j.simpa.2021.100073

Abstract

MinViME is a python package and web application that allows scientists and engineers to estimate a lower bound on the performance characteristics of a machine learning model for a problem with well quantified characteristics. It is used to evaluate the feasibility of machine learning projects, prioritize among multiple competing projects, and run simulations to explore the structure of the difficulty landscape in a multi-dimensional space of potential problems.

Download paper here