A herd of heuristic algorithms is compared using a portfolio optimization. Previously “A comparison of some heuristic optimization methods” used two
Previously “A comparison of some heuristic optimization methods” used two simple and tiny portfolio optimization problems to compare a number of optimization functions in the R language. This post expands upon that by using a portfolio optimization problem that is of a realistic size (but still with an unrealistic lack of constraints). Test case The optimization problem is to select 30 assets out of a universe of 474 and find their best weights. The weights must be non-negative and sum to 1. The integer constraint is binding — more than 30 assets in the portfolio can give a better utility. The utility is mean-variance. Each optimizer was run 100 times. To have a fair comparison the amount of time that each run took was controlled to be about 1 kilosecond. (There are a couple that also have runs that take much less time.) The timings are not all particularly close to 1000 seconds, but they are probably all close enough that the picture is minimally distorted.