Deep learning for NeuroImaging in Python.

Note

This page is a reference documentation. It only explains the class signature, and not how to use it. Please refer to the gallery for the big picture.

class surfify.nn.modules.IcoDiNeConv(in_feats, out_feats, neigh_indices, bias=True)[source]¶

The convolutional layer on icosahedron discretized sphere using n-ring filter (based on the Direct Neighbor (DiNe) formulation).

See also

IcoRePaConv, IcoSpMaConv

Notes

Debuging messages can be displayed by changing the log level using setup_logging(level='debug').

Examples

>>> import torch
>>> from surfify.nn import IcoDiNeConv
>>> from surfify.utils import icosahedron, neighbors
>>> ico2_vertices, ico2_triangles = icosahedron(order=2)
>>> neighbor_indices = neighbors(
        ico2_vertices, ico2_triangles, depth=1, direct_neighbor=True)
>>> neighbor_indices = np.asarray(list(neighbor_indices.values()))
>>> module = IcoDiNeConv(
        in_feats=8, out_feats=8, neigh_indices=neighbor_indices)
>>> ico2_x = torch.zeros((10, 8, len(ico2_vertices)))
>>> ico2_x = module(ico2_x)
>>> ico2_x.shape

Init IcoDiNeConv.

Parameters:

in_feats : int

input features/channels.

out_feats : int

output features/channels.

neigh_indices : array (N, k)

conv layer’s filters’ neighborhood indices, where N is the ico number of vertices and k the considered nodes neighbors.

bias : bool, default True

the layer will learn / not learn an additive bias.

forward(x)[source]¶: Forward method.