NitroML is a modular, portable, and scalable model-quality benchmarking framework for Machine Learning and Automated Machine Learning (AutoML) pipelines.
Introduce MetaFeatureGen component that is used to create meta-features for a given dataset.
Introduce MetaLearner component that recommends a HParam configuration based on majority voting algorithm. The component takes candidate HParam configurations learned using multiple training datasets.
A meta-learning pipeline example that constructs an end-to-end meta-learning DAG.
MetaFeatureGen
component that is used to create meta-features for a given dataset.MetaLearner
component that recommends a HParam configuration based on majority voting algorithm. The component takes candidate HParam configurations learned using multiple training datasets.