scikit-elm is a scikit-learn compatible Extreme Learning Machine (ELM) regressor/classifier.
It features very high degree of model flexibility: dynamically added classes,
partial_fit without performance penalties, wide data format compatibility,
optimization and parameter selection without full re-training.
Big Data and out-of-core learning support through dask-powered backend. GPU acceleration support with NVidia hardware, and on macOS through plaidml.
Toolbox is in active development, initial release soon.