Potential Bug in iter_sparse: Possible Missing para (paral1) Multiplication vs. PDF Derivation

[iljgl]

Hi team,

疑似bug_SparseDeconvolution.pdf

I’ve been reviewing the iter_sparse function as part of the SparseHessian_core workflow, and noticed a potential inconsistency between its implementation and the sparse optimization logic derived in the attached PDF. I’m sharing this to confirm if it was a real Bug, as my deduction may be wrong.

[WeisongZhao]

Hi,

Yes, you're right. Actually, there have been several versions of this method, and the one we've shared publicly is the most stable under a wide range of parameter settings. There's always a gap between what's mathematically optimal and what runs robustly in practice.

Feel free to tweak or adapt the code to better suit your needs!

Hope this helps,
Weisong