Please check the Store Optimization Report in File -> Preferences -> Workspaces -> Saving Data.
Very good. That solves that. Thanks!
I have another observation.
In the Exhaustive Search Properties dialog it seems that when a box is UNchecked (meaning do not optimize this param, and ignore the range settings without perturbing them) the opt runs not with the Current Value (as it should), but with the Start Value of that param.
Then, after the opt, if you go to apply one of the opt results, the Current Value is back in effect, meaning that you cannot get the result that you see from the opt table instantiated on the chart.
Correct behavior would be:
When a param box is unchecked, use the Current Value during the opt
When a param box is checked, use the Start Value, End Value, and Step Value during the opt
In all cases, when applying an opt report result, the chart should take the params used in that iteration and thus agree with the opt report results.
If I am misunderstanding the current facts, please let me know. Thanks.
MC_Prog,
Expected behavior is the following:
When an input box is unchecked, use the Start Value during the optimization. When you apply the optimization results, Current Value becomes equal to Start Value used during the optimization.
This makes it not possible to perform the proper guidance of sequential optimizations by using the checkboxes (which is the point of them in the first place).
The as-is design requires the user to manually type the Current Value into the Start Value column after each step of the sequential optimization.
Correct behavior (design), as I wrote above, would keep good found values without requiring them to be retyped by the user. Instead, just uncheck them!
Please consider the use case:
1. Default values are set to "something thought maybe reasonable"
2. Inputs related to Entry are optimized (by checking them)
3. A good performance iteration is selected from the opt results. (choosing desired Entry Inputs)
4. Inputs related to Exit are now optimized (by checking them, with Entry Inputs unchecked)
5. A good performance iteration is selected from the opt results. (choosing desired Exit Inputs)
We now have what is typically a pretty darn good result, and it took (for example) 2000 iterations to arrive at rather than 1 million. (Aside: it is true we did not cover the whole search space, but that is exactly the point!!)
In my use case, no parameters had to be retyped at all, ever. In my use case, there is never a disagreement between where the user thinks they started (parameter-wise) and where he ends up. In my use case, no result is ever lost and every step of sequential optimization will yield better (or at very worst equal) results. In my use case, the chart will always agree with a selected optimization result.
With respect for the efforts expended so far, the currently expected behavior does need an upgrade!
We are going to be living with 8.5 in the field for a long time. For those of us doing alot of optimzation, it makes a
huge difference whether or not the checkbox feature accomplishes the use case or not. Please consider retooling this one more time before going final.
