ttclass/amensum.m

Tensor train summation without loss of accuracy.

Syntax

    y=amensum(x,tol,opts)


Description

Sums buffered tensor trains in a single tensor train using AMEn algorithm.

Arguments

   x   - ttclass with buffered rank-one tensors

tol - relative tolerance parameter, e.g. 1e-10

The opts field is optional:

opts.max_swp - maximum number of iterations

opts.init_guess_rank - rank of the initial guess

opts.enrichment_rank - rank of the enrichment

opts.verb            - verbosity switch


Outputs

   y   - ttclass with a single tensor train, such
that |x-y|<tol*|x| in Frobenius norm


Notes

Summation is a messy operation in TT format. Spinach buffers tensor train sums and processes them in batches using this function.