ARD EM algorithm with automatic determination of components/clusters number
# ARD EM ARD (Automatic Relevance Determination) EM implementation on Python. The classical EM-algorithm for reconstructing a mixture of normal distributions does not allow to determine the amount of components of the mixture. The ARD EM implementation suggests algorithm for automatically determining the number of components ARD EM, based on the method of relevant vectors. The idea of the algorithm is to use at the initial stage of a knowingly excessive amount of the components of the mixture with further determination of the relevant components by maximizing validity. Experiments on model problems show that the number of found clusters either coincides with the true one, or slightly excels him. In addition, clustering with ARD EM is closer to the true than the analogs based on sliding control and character of the minimum description length. It’s EM algorithm with automatic determination of number of components. It’s powerful and fast algorithm for gaussian mixture learning and clustering with unknown number of components.
# Implementation The implemented [GaussianMixtureARD](ard_em.py) class has the same interface as SkLearn’s [GaussianMixture](http://scikit-learn.org/stable/modules/generated/sklearn.mixture.GaussianMixture.html#sklearn.mixture.GaussianMixture) one, but with 3 additional parameters: `python init_components="sqrt" # Initial number of components. sqrt(N) if "sqrt" alpha_bound=1e3 # Drop all components with weight_reg (alpha) > alpha_bound weight_bound=1e-3 # Drop all components with weight < weight_bound ` and without n_components one.
# Installation ` pip install git+https://github.com/Leensman/ard-em.git `
## Example `python from ard_em import GaussianMixtureARD gmm = GaussianMixtureARD() gmm = gmm.fit(X) print('Bayesian information criterion: ', gmm.bic(X)) best_n_components = gmm.n_components print('Best number of components: ', best_n_components) gmm.predict(X) ` For more examples go to [GaussianMixture.ipynb](https://github.com/Leensman/ard-em/blob/master/ard-em/examples/Gaussian%20mixture.ipynb)
## Links [Original paper](http://www.machinelearning.ru/wiki/images/d/dc/Vetrov-ArdEm-JVMMF-2009.pdf)
## Author Artem Ryzhikov, LAMBDA laboratory, Higher School of Economics, Yandex School of Data Analysis
HSE profile: https://www.hse.ru/org/persons/190912317
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size ard_em-0.1.4-py3.6.egg (8.5 kB)||File type Egg||Python version 3.6||Upload date||Hashes View hashes|
|Filename, size ard_em-0.1.4-py3-none-any.whl (7.2 kB)||File type Wheel||Python version py3||Upload date||Hashes View hashes|
|Filename, size ard_em-0.1.4.tar.gz (5.7 kB)||File type Source||Python version None||Upload date||Hashes View hashes|