dimension module

btQuant.dimension.ica(X, nComponents=None, maxIter=200, tol=0.0001)[source]

Independent component analysis.

Parameters:

  • X – features

  • nComponents – number of components

  • maxIter – maximum iterations

  • tol – convergence tolerance

Returns:

dict with ‘transformed’, ‘mixingMatrix’, ‘unmixingMatrix’

btQuant.dimension.isomap(X, nComponents=2, nNeighbors=5)[source]

Isomap non-linear dimensionality reduction.

Parameters:

  • X – features

  • nComponents – target dimensions

  • nNeighbors – number of nearest neighbors

Returns:

dict with ‘transformed’

btQuant.dimension.kernelPca(X, nComponents=None, kernel='rbf', gamma=None)[source]

Kernel PCA for non-linear dimensionality reduction.

Parameters:

  • X – features

  • nComponents – number of components

  • kernel – ‘rbf’, ‘poly’, ‘linear’

  • gamma – kernel parameter

Returns:

dict with ‘transformed’, ‘eigenvalues’, ‘eigenvectors’

btQuant.dimension.lda(X, y, nComponents=None)[source]

Linear discriminant analysis.

Parameters:

  • X – features

  • y – class labels

  • nComponents – number of components

Returns:

dict with ‘transformed’, ‘eigenvalues’, ‘eigenvectors’

btQuant.dimension.mds(X, nComponents=2, metric=True, maxIter=300)[source]

Multidimensional scaling.

Parameters:

  • X – features or distance matrix

  • nComponents – target dimensions

  • metric – use metric MDS (True) or non-metric (False)

  • maxIter – maximum iterations for non-metric

Returns:

dict with ‘transformed’

btQuant.dimension.nmf(X, nComponents, maxIter=200, tol=0.0001)[source]

Non-negative matrix factorization.

Parameters:

  • X – non-negative features

  • nComponents – number of components

  • maxIter – maximum iterations

  • tol – convergence tolerance

Returns:

dict with ‘W’ (basis), ‘H’ (coefficients)

btQuant.dimension.pca(X, nComponents=None)[source]

Principal component analysis.

Parameters:

  • X – features (nSamples x nFeatures)

  • nComponents – number of components to keep

Returns:

dict with ‘transformed’, ‘eigenvalues’, ‘eigenvectors’, ‘explained_variance’

btQuant.dimension.tsne(X, nComponents=2, perplexity=30.0, maxIter=1000, learningRate=200.0)[source]

t-SNE dimensionality reduction.

Parameters:

  • X – features

  • nComponents – target dimensions

  • perplexity – perplexity parameter

  • maxIter – maximum iterations

  • learningRate – learning rate

Returns:

dict with ‘transformed’