morphoclass.unsupervised package

Module contents

Utilities for unsupervised methods for neuron m-type classification.

morphoclass.unsupervised.make_silhouette_plot(embeddings, n_clusters, random_state=0)

Create a silhouette plot given an embedding.

This code up to small amendments is copied from https://scikit-learn.org/stable/auto_examples/cluster/plot_kmeans_silhouette_analysis.html

Parameters

  • embeddings (np.ndarray) – The embedding to be plotted. The shape should be (n_samples, d_embedding).

  • n_clusters (int) – The number of clusters to create.

  • random_state (int) – The random state to use for KMeans (first plot) and PCA (second plot).

morphoclass.unsupervised.plot_embedding_pca(embeddings, ax=None, pca=None, labels=None, title=None, annotate_indices=False, annotation_fontsize=7)

Plot two principal components of given embedding.

This is an altered version which shows annotations.

Parameters

  • embeddings (np.ndarray) – The embedding to be plotted. The shape should be (n_samples, d_embedding).

  • ax (matplotlib.axes.Axes, optional) – A matplotlib axis object.

  • pca (sklearn.decomposition.PCA, optional) – An fitted instance of the PCA class from scikit-learn. If not provided then a new instance will be created and fitted.

  • labels (list_like, optional) – Labels for all samples. Should be of length n_samples.

  • title (str) – Title for the plot.

  • annotate_indices (bool) – If true then each point in the plot will be annotated with its index in the provided embeddings.

  • annotation_fontsize (int) – The font size for the annotation. Has no effect if annotate_indices is False.

Returns

pca – The PCA of the given embeddings.

Return type

sklearn.decomposition.PCA