mlreco.models.layers.gnn.cluster module

mlreco.models.layers.gnn.cluster.gauss(centroid, sigma)[source]

Constructor for a gaussian kernel functions.

INPUTS:

  • centroid: (D, ) Tensor for the coordinates of the gaussian centroid.

  • sigma: value for gaussian bandwidth.

Returns

kernel function defined by centroid and sigma.

Return type

  • f (function)

mlreco.models.layers.gnn.cluster.mvgauss(centroid, L, dim=3)[source]

Constructor for multivariate gaussian kernels.

L (torch.Tensor): D x D tensor representing Cholesky decomposition of the covariance matrix. The covariance matrix is then calculated as:

Sigma = LL^T.

mlreco.models.layers.gnn.cluster.laplace(centroid, sigma)[source]
mlreco.models.layers.gnn.cluster.student_t(centroid)[source]

Pairwise student t distribution as used in TSNE

mlreco.models.layers.gnn.message_passing.nnconv_old module mlreco.models.layers.gnn.factories module