Fred's Intelligent Optimizer (IO) in the file section, is an add-on product that provides a non-exhaustive optimization capability for systems with many parameters. One of the features of IO is the sensitivity adjusted optimization. It has a way to avoid the peaks by also testing the surrounding data points in the N-dimensional solution space. IO let's you control how much of your precious CPU time is spent on sensitivity testing. It also produces graphs of the sensitivity of the overall solution as well as the sensitivity of each parameter by itself.
-Steve --- In [email protected], "Tomasz Lutelmowski" <[EMAIL PROTECTED]> wrote: > > Hello, > > I found optimizing with one or two parameters is quite easy. For example, if > I want to optimize two parameters I look at 3D optimization graph, then look > for large areas ("high level islands" but ignoring peaks at the same time) > where my optimization target is relatively high, and use coordinates of > middle of these islands as optimized parameters. Optimizing this way I hope > to be on safe side and not over-adjust system parameters. > > I suppose that there is a way of not taking this results at face value, and > maybe it is the only way of optimizing systems with more than 2 parameters. > I think derivatives and convex function might be somehow applicable, but I > have no clue of how to implement it. Especially if I want to use BackTester > results lists as input, not the function. So my question is - do you know > any method of reliable estimation of optimization parameters for systems > with 3-5 of them ? > > Best regards, > Tomek >
