sb-ai-lab/LightAutoML

What it solves

LightAutoML (LAMA) simplifies the process of creating machine learning models by automating the pipeline construction. It removes the need for manual feature engineering and hyperparameter tuning, allowing users to create high-performing models with minimal code.

How it works

LAMA provides two primary modes of operation:

1. Ready-to-use presets: Users can apply pre-configured pipelines (e.g., TabularAutoML) for common tasks like binary classification or regression on tabular, time series, image, and text data.
2. Framework mode: For advanced users, LAMA acts as a framework where they can build custom ML pipelines using a library of ready-to-use blocks, including components for feature generation, selection, and hyperparameter optimization.

Who it’s for

It is designed for data scientists and ML engineers who want to rapidly prototype models or build complex, customized ML pipelines without starting from scratch.

Highlights

• Multi-modal support: Handles tabular, time series, image, and text data.
• Extensible architecture: Offers both high-level presets for quick starts and a low-level framework for custom pipeline construction.
• Interpretability: Includes tools for model interpretation, such as WhiteBox, ICE, and PDP approaches.
• Broad integration: Supports data sources from CSVs and SQL databases.