cheap_norm.m

The cheapest norm for various representations of matrices. CUDA stores matrices by rows, Matlab by columns, and polyadic objects can only multiply vectors, in which case Algorithm 2.4 from Higham and Tisseur's paper (https://doi.org/10.1137/S0895479899356080) is used.

Syntax

    n=cheap_norm(A,t,itmax)

Parameters

    A     - a matrix, or a polyadic representation thereof

    t     - optional number of probe columns in the polyadic norm
            estimator, defaults to 1

    itmax - optional maximum number of estimator iterations,
            defaults to 5

Outputs

    n - infinity-norm for GPU arrays, 1-norm for CPU arrays,
        and a lower-bound 1-norm estimate for polyadics

Notes

Some norms are vastly more expensive than others; this function uses the cheapest ones available.

See also

clean_up.m, frob_chop.m, hdot.m, mprealloc.m, acomm.m, arnoldi.m, atranspose.m, aux_mat.m, binpack.m, cheb_coeff.m, dirdiff.m, eigenfields.m, expdrop.m, expmint.m, expmint2.m, fftdiff.m, fourdif.m, fourlap.m, gaussfun.m, herm_spline.m, jacobianest.m, keep_rank.m, krondelta.m, kronm_new.m, logfactorial.m, lorentzcon.m, lorentzfun.m, md5_hash.m, remncomm.m, remtrace.m, rspert.m, rspt_eig.m, snormpdf.m, svd_shrink.m, tikhoind.m, tikhonov.m, trapdiff.m, unit_oper.m, unit_state.m, vvpert.m, Kernel_utilities

Version 2.3, authors: Ilya Kuprov