Fork me on GitHub
#data-science
<
2018-07-12
>
blueberry00:07:46

@justalanm Small update about eigen performance: In numpy code, you're using the eigh function, which assumes your matrix is symmetric. That, of course, is faster than computing the non*symmetric eigenvalue problem in neanderthal code. The current version of neanderthal has many specialized polymorphic calls for symmetric matrices. Unfortunately, not yet for symmetric eigenproblems. I'll add that in the next version. Generally, if you know that you have a matrix that is symmetric, use *sy (or view-sy), not ge matrices.

alan05:07:24

@blueberry I'm slowly adding corrections (not much time right now unfortunately) how would I create a SY in the most efficient way?

blueberry07:07:40

call view-sy if you already have a matrix with data, or if you want it to be sy from the beginning, call sy instead of ge.