airbnb/aerosolve

Aerosolve is a machine learning library that provides a comprehensive framework for training, evaluating, and scoring predictive models. It supports a wide variety of algorithms, including linear models, decision trees, random forests, and neural networks, while offering extensive feature engineering and transformation capabilities. The system is designed to integrate with Apache Spark for distributed processing and includes utilities for data conversion and integration with other ML platforms like PhotonML.

This period marks the initial release of the Aerosolve project, establishing its foundational build system, CI/CD infrastructure, and core machine learning capabilities. It introduces a comprehensive set of feature extraction, model implementations, and data transformation utilities, alongside extensive unit tests and demo applications for image processing and income prediction.

20 changes

This period focused on establishing a unified training and evaluation framework, introducing a generic pipeline for model training, scoring, and evaluation. Significant work was also done on core feature engineering, adding new function types and configuration classes to support flexible model building. Additionally, a demo for the Twenty News dataset was created to demonstrate the new capabilities.

17 changes

This period focused on expanding the Airlearner module with new binary regression strategies and XGBoost parameter search capabilities. The work included adding comprehensive documentation, utility libraries for data engineering, and extensive test coverage to support these new features.

5 changes

airbnb/aerosolve was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point the surveyor at a repository you know and see whether you agree with it.