A simple, extensible backend for developing auto-tuning systems
Overview¶
License: MIT
Development Status: Pre-Alpha
Documentation: https://MLBazaar.github.io/BTB
Homepage: https://github.com/MLBazaar/BTB
BTB (“Bayesian Tuning and Bandits”) is a simple, extensible backend for developing auto-tuning systems such as AutoML systems. It provides an easy-to-use interface for tuning and selection.
It is currently being used in several AutoML systems:
ATM, distributed, multi-tenant AutoML system for classifier tuning
mit-d3m-ta2, MIT’s system for the DARPA Data-driven discovery of models (D3M) program
AutoBazaar, a flexible, general-purpose AutoML system
History¶
In its first iteration, in 2018, BTB was designed as an open source library that handles the problems of tuning the hyperparameters of a machine learning pipeline, selecting between multiple pipelines and recommending a pipeline. A good reference to see our design rationale at that time is Laura Gustafson’s thesis, written under the supervision of Kalyan Veeramachaneni:
Bayesian Tuning and Bandits. Laura Gustafson. Masters thesis, MIT EECS, 2018.
Later in 2018, Carles Sala joined the project to make it grow as a reliable open-source library that would become part of a bigger software ecosystem designed to facilitate the development of robust end-to-end solutions based on Machine Learning tools. This second iteration of our work was presented in 2019 as part of the Machine Learning Bazaar:
The Machine Learning Bazaar: Harnessing the ML Ecosystem for Effective System Development. Micah J. Smith, Carles Sala, James Max Kanter, and Kalyan Veeramachaneni. Sigmod 2020.