Question on code v.s. FASTA field guide documentation

[lequant40]

Hello !

Looking at

fasta-matlab/solvers/fasta.m

Line 154 in 033c41b

residual(i) = norm(Dx(:))/tau0; % Estimate of the gradient, should be zero at solution

v.s. FASTA field guide (https://arxiv.org/pdf/1411.3406.pdf), section 4.6 Stopping Conditions, I would have a question/remark/issue.

Looking at the variable normalizer, I find it to be the same as the denominator in formula 42 of the FASTA field guide:

fasta-matlab/solvers/fasta.m

Line 156 in 033c41b

normalizer = max(norm(gradf0(:)),norm(x1(:)-x1hat(:))/tau0)+opts.eps_n;

Now, looking at the variable residual(i) (which, together with variable normalizer makes for the full formula 42 of the FASTA field guide), I fail to understand why it is equal to norm(Dx(:))/tau0 = norm(x1(:) - x0(:))/tau0 and not to norm(gradf0(:) + (x1(:)-x1hat(:))/tau0) ?

Cheers,

Roman

PS: By the way, your FASTA field guide greatly helped me to implement my own twists on FISTA, thanks a lot !