dodiscover.ci.kernel_utils.compute_kernel#
- dodiscover.ci.kernel_utils.compute_kernel(X, Y=None, metric='rbf', distance_metric='euclidean', kwidth=None, centered=True, n_jobs=None)[source]#
Compute a kernel matrix and corresponding width.
Also optionally estimates the kernel width parameter.
- Parameters:
- Xarray_like of shape (n_samples_X, n_features_X)
The X array.
- Yarray_like of shape (n_samples_Y, n_features_Y), optional
The Y array, by default None.
- metric
str, optional The metric to compute the kernel function, by default ‘rbf’. Can be any string as defined in
sklearn.metrics.pairwise.pairwise_kernels(). Note ‘rbf’ and ‘gaussian’ are the same metric.- distance_metric
str, optional The distance metric to compute distances among samples within each data matrix, by default ‘euclidean’. Can be any valid string as defined in
sklearn.metrics.pairwise_distances().- kwidth
float, optional The kernel width, by default None, which will then be estimated as the median L2 distance between the X features.
- centeredbool, optional
Whether to center the kernel matrix or not, by default True.
- n_jobs
int, optional The number of jobs to run computations in parallel, by default None.
- Returns:
- kernelarray_like of shape (n_samples_X, n_samples_X) or (n_samples_X, n_samples_Y)
The kernel matrix.
- med
float The estimated kernel width.