As a user, I would like to be able to convert the selected indices of a TensorTable into a "Sparse" or linearized TensorTable that could later be used to update the TensorTable. Conversely, I would like to be able to reconstruct/update the tensorTable from a "Sparse" TensorTable.
Objectives
[x] Method to convert from a TensorTable to a Sparse representation
[x] Method to be able to update a TensorTable from a Sparse representation
[x] Unit tests on all new methods for CPU/GPU versions
Validaiton
[x] Passing unit tests
Details
getSelectionAsSparseTensorTable
Create a 2D TensorTable with axes for the indices and axis for the values
The indices axis dimensions size = TDim
The indices labels are of type int and are the selected indices values
updateTensorTableFromSparseTensorTable
Select the TensorTable based off of the indices in the indices axis of the sparse Tensor Table
Sort the TensorTable according to the selected values in "Column-wise" order to align the values of the "Sparse" TensorTable with the "Dense" TensorTable.
Update the 1D TensorTable slice with the sparse values
Description
As a user, I would like to be able to convert the selected indices of a TensorTable into a "Sparse" or linearized TensorTable that could later be used to update the TensorTable. Conversely, I would like to be able to reconstruct/update the tensorTable from a "Sparse" TensorTable.
Objectives
Validaiton
Details
getSelectionAsSparseTensorTable
intand are the selected indices valuesupdateTensorTableFromSparseTensorTable