Visualization

repliclust.viz

Provides the built-in visualization features of repliclust.

Functions:

plot()

Plot a dataset.

repliclust.viz.plot(X, y=None, dimensionality_reduction='tsne', dim_red_params={}, **plot_params)

Plot high-dimensional data with dimensionality reduction and clustering labels.

This function creates a 2D scatter plot of the input data X. If X has more than two features, dimensionality reduction is performed using either t-SNE or UMAP before plotting. Optionally, data points can be colored according to cluster labels provided in y.

Parameters:

• X (array-like of shape (n_samples, n_features)) – The input data to plot.

• y (array-like of shape (n_samples,), optional) – Cluster labels or target values used to color the data points. If None, all points are plotted with the same color.

• dimensionality_reduction ({'tsne', 'umap'}, default='tsne') –

  The method used for dimensionality reduction when X has more than two features. Choices are:

  • ’tsne’ : Use t-distributed Stochastic Neighbor Embedding.

  • ’umap’ : Use Uniform Manifold Approximation and Projection.

• dim_red_params (dict, default={}) – Additional keyword arguments to pass to the dimensionality reduction algorithm.

• **plot_params – Additional keyword arguments passed to matplotlib.pyplot.scatter.

Raises:

• ValueError – If X has fewer than two features.

• ValueError – If dimensionality_reduction is not one of ‘tsne’ or ‘umap’.

See also

matplotlib.pyplot.scatter

Create a scatter plot.

sklearn.manifold.TSNE

t-distributed Stochastic Neighbor Embedding.

umap.UMAP

Uniform Manifold Approximation and Projection.

Examples

Plot data with t-SNE dimensionality reduction:

>>> plot(X, y, dimensionality_reduction='tsne')

Plot data with UMAP dimensionality reduction and custom parameters:

>>> dim_red_params = {'n_neighbors': 15, 'min_dist': 0.1}
>>> plot(X, y, dimensionality_reduction='umap', dim_red_params=dim_red_params)

Plot 2D data without dimensionality reduction:

>>> X_2d = np.random.rand(100, 2)
>>> plot(X_2d, y)