With the recent release of version 1.0 in March 2022, FLAML, Microsoft’s open source AutoML library takes a big leap towards productive usage.

Fast progress on fast AutoML

When FLAML started in late 2020, the first release includes hyperparameter and sample size methods. For example, developers added “cost frugal optimization” (CFO) and “estimated cost of improvement” (ECI) early. The github source already features a notebook with demo code on how to run FLAML. Since then, another 33 releases were added to the repository, including a variety of options to make AutoML adaptable and ease to use by people with several levels of coding skills. Microsoft’s addition to the AutoML toolstack is one of the reasons why even more users are able to benefit from AutoML – without requiring them to have a PhD.

What makes FLAML’s v1.0 update a key milestone?

To give a deeper view on where FLAML stands today, let’s dive into technical details. FLAML developers @github/sonichi, @github/qingyun-wu et al. released a new feature called “zero-shot AutoML” already in v0.10.0. These were key contributions from the Microsoft team to reach the important v1.0 step. In short, they brought autoML features to existing, widely used libraries like xgboost and others.
Let me cite from the changelog: “This release contains an important new feature: zero-shot AutoML and mete learning. It provides a new way of doing AutoML without tuning. You can now use the existing training API from lightgbm, xgboost etc. while getting the benefit of AutoML in choosing high-performance hyperparameter configurations per task. Recommended for everyone currently using lightgbm, xgboost or random forest, regardless of previous experience in AutoML.”

FLAML’s relevance for AutoML

At bergamot_ai, we watch this development with excitement. The decision to give FLAML to the community with an MIT based license means others can make use of it very flexibly. There’s a good chance the one or another FLAML feature is incorporated into bergamot_ai the future, provided that the library continues to be under active development and sparks further interest in the open source community.

How you can start with AutoML as a non-coder

In case you’re unsure whether FLAML is for you, we would recommend to try it out yourself. If you are unsure how to benefit from AutoML methods to tackle your data challenges without the hard coding part, do not worry. there are tools around already. Read about our contribution bergamot_ai here. Or feel free to contact us directly via the contact form.

Start your predictive analytics use case in minutes with bergamot_ai. Get the demo today.