Sapientml

SapientML is an AutoML library that generates machine learning pipelines for predictive tasks. It learns from a corpus of existing datasets and the human written pipelines built for them, then uses that knowledge to produce a high quality pipeline for a new dataset instead of searching every possible combination. It is aimed at data scientists and engineers who want to build predictive models quickly while still seeing how the model was made. The site names three strengths. It is fast because it evaluates only the most plausible pipelines. It is transparent because the output is a generated machine learning program that contains explanations, which users can read and edit. It is accurate because its predictions draw on the programs previously used to build accurate models. Getting started takes a pip install of the sapientml package followed by calls to its APIs to generate pipelines, with a getting started guide in the documentation. The approach was published as a paper at the 44th International Conference on Software Engineering, ICSE 2022, titled Synthesizing Machine Learning Pipelines by Learning from Human Written Solutions, and the site links to the preprint. The site publishes no pricing and presents the library as open source tooling with no paid plans.