speed concerns

[quaquel]

I am playing around with platypus and the inter temporal lake problem. I have implemented a generational version of BORG using MultiMethod and NSGA2. I however now run into severe run problems. For example 60% of my runtime is spend in the orthogonalize function. A quick look at the code easily explain why: everything is done using lists.

Hence my question: are there specific reasons for not using numpy? A simple implementation of orthogonalize using numpy with the inter temporal lake problem (so 100 decision variables) is a factor 10 faster.