andymilne/Sparse-Array-Toolbox

For sparse data, sparse formats can significantly increase speed and reduce memory requirements. However, MATLAB’s built-in sparse format represents only vectors and matrices, not N-dimensional arrays (tensors). This toolbox contains a new "sparse array structure" and associated operations – outer, entrywise, and inner products, addition, summation, convolution, distance measures, permutation, (circular) shifts, and other manipulations – that can be applied to N-dimensional full arrays or to sparse array structures. All functions have been carefully designed to optimize speed.
See SparseArrayToolbox.pdf for more information.