Oily tensor

Question

Does Matlab support efficient operations on large sparse tensors?

More specific:

• Is there an elegant way, like sparse , loading and storing a sparse tensor? As far as I understand, sparse can only load matrices.
• Are operations like a tensor product effective on sparse tensors?

I understand that I can always store the tensor as a combination of arrays of matrix cells, but this will require the use of loops, and I hope to avoid this.

Since the data I work with is very large, I cannot consider a non-sparse representation.

Answer 1

Out of the box, I believe that MATLAB only processes sparse matrices, as you say.

But you can take a look at the Tensor Toolbox and the N-way Toolbox to see if they fit your needs. Both are freely available, and I heard good things about everyone (although I didn't use it myself). Tensor Toolbox, in particular, apparently has at least some support for sparse multidimensional arrays.

Answer 2

You can use the Tensor Toolbox to work with tensors. you can use sptensor () to create a sparse tensor in this toolbar.

Answer 3

If you are looking for a truly scalable solution, take a look at SPLATT: http://glaros.dtc.umn.edu/gkhome/splatt/overview