Runs one evaluation against the full epoch of data. Return the precision and the number of correct predictions. Batch evaluat

Restore parameters if no session given.

Build the computational graph of the model.

Return a list of indices to reorder the adjacency and data matrices so that the union of two neighbors from layer to layer forms a binary tre

Fully connected layer with Mout features.

Fully connected layer with Mout features.

Keep the documents given by the index, discard the others.

Lanczos algorithm which computes the orthogonal matrix V and the tri-diagonal matrix H.

Return the B-spline basis. K: number of control points. x: evaluation points or number of evenly distributed evaluation points.

Diagonalize the tri-diagonal H matrix.

Return the Fourier basis, i.e. the EVD of the Laplacian.

It builds the model, i.e. the computational graph, as far as is required for running the network forward to make predictions,

Keep the documents given by the index, discard the others.

Adds to the inference model the layers required to generate loss.

Coarsen a graph multiple times using the METIS algorithm. INPUT W: symmetric sparse weight (adjacency) matrix levels: the number of

Permute adjacency matrix, i.e. exchange node ids, so that binary unions form the clustering tree.

Permute data matrix, i.e. exchange node ids, so that binary unions form the clustering tree.

Return a simplified version of this mesh. A Qslim-style approach is used here. :param factor: fraction of the original vertices to retain

Adds to the loss model the Ops required to generate and apply gradients.

Unpermute data back to its original form

Computes a quadric for each vertex in the Mesh. Returns: v_quadrics: an (N x 4 x 4) array, where N is # vertices.

Return the adjacency matrix of a kNN graph.

Average pooling of size p. Should be a power of 2.

Bias and ReLU. One bias per filter.

Bias and ReLU. One bias per filter.

Bias and ReLU. One bias per vertex per filter.