Module: tfg.geometry.convolution.utils

View source on GitHub

This module implements various sparse data utilities for graphs and meshes.

Functions

check_valid_graph_convolution_input(...): Checks that the inputs are valid for graph convolution ops.

check_valid_graph_pooling_input(...): Checks that the inputs are valid for graph pooling.

check_valid_graph_unpooling_input(...): Checks that the inputs are valid for graph unpooling.

convert_to_block_diag_2d(...): Convert a batch of 2d SparseTensors to a 2d block diagonal SparseTensor.

flatten_batch_to_2d(...): Reshapes a batch of 2d Tensors by flattening across the batch dimensions.

partition_sums_2d(...): Sum over subsets of rows in a 2-D tensor.

unflatten_2d_to_batch(...): Reshapes a 2d Tensor into a batch of 2d Tensors.