0

How do I make non-linear portfolio optimization faster? I want to optimize a portfolio of 50 assets on the following measure: (mean(portfolio returns_t-1)/sd(portfolio returns_t-1)-mean(portfolio returns_t-2)/sd(portfolio returns_t-2). So I want the measure being as high as possible by choosing according portfolio weights

Currently I use the function DEoptim on R, however, with 50 assets the running time is quite long and I don’t know whether the solution is globally optimal or only locally. Is there a solution to make portfolio optimization on such difficult measures faster? Better algorithms? Switching to different programming languages? Thanks in advance!

Lukeseb93
  • 1
  • 2
  • Please check [how to ask](https://stackoverflow.com/help/how-to-ask) and provide a [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) – Emmanuel-Lin Dec 05 '17 at 12:17
  • I don't understand your objective function but in general for smooth NLP (nonlinear programming) problems a real NLP solver is more appropriate than a DE heuristic (which is basically "trial-and-error"). – Erwin Kalvelagen Dec 05 '17 at 12:32

0 Answers0