/hpelm

High performance implementation of Extreme Learning Machines (fast randomized neural networks).

Primary LanguageC

High Performance ELM

Extreme Learning Machine (ELM) with model selection and regularizations.

In-memory ELM works, check hpelm/tests folder. MAGMA acceleration works, check hpelm/acc/setup_gpu.py.

Example usage:

>>> from hpelm import ELM
>>> elm = ELM(X.shape[1], T.shape[1])
>>> elm.add_neurons(20, "sigm")
>>> elm.add_neurons(10, "rbf_l2")
>>> elm.train(X, T, "LOO")
>>> Y = elm.predict(X)

If you use the toolbox, cite our paper "High Performance Extreme Learning Machines: A Complete Toolbox for Big Data Applications" that will be published in IEEE Access.