Here are my initial comments. Some may sound critical but they are meant to be constructive and Iām sure others will probably have opposing views.
(1) The total of all nodes add up to 1 in this version, but there are only 2 significant digits. I feel that the weighting resolution is too coarse. I would like to see 3 significant digits (at least). One of my longer term objectives is to use the tool to create ātest vectorsā or test RSās. I would start with a production ranking system but have a series of RSās with minor amounts of random noise applied to each node weight (up to 5%). The noisy RSās would be applied against the production port/simulation to see how the sim responds. This simply would not be possible with the optimizer as is. But adding a third decimal place would increase the size of the spreadsheet horizontally (see problem #2). So I am wondering if the sum of nodes could be re-scaled to 10 instead of 1. This would not take up any additional screen space.
(2) I find that the horizontal and vertical size of the spreadsheet is too large. Now perhaps I have a Windows setup problem, I donāt know. But Iād like to explain that I use two monitors, the first is the screen attached to my portable PC. The second is a separate 23" widescreen display. The two monitors are very convenient for using this ranking system optimizer because I can have both P123 and the spreadsheet displayed at the same time. So the problem is that with two monitors selected, Windows adjusts the effective resolution of each graphics display to be the same. I donāt get more detail on the 23" monitor than the portable PC display. I have attached what I see when I open up the spreadsheet. The active area is way too large (for me) as I donāt like messing around with scroll bars, particularly when I am doing a very repetitive job. The other issue is that even with my old spreadsheet (10 iterations), I would not be able to complete my work without getting a āMax number of requests - 500 in 6 hoursā. The 20 iterations will make that cap come twice as fast I think what you have is good for newbies, but I am wondering if in the end a stripped down version could be made (with no training wheels), perhaps 15x15, with the preamble at the top shrunk/minimized and fit onto the side instead of the top of the spreadsheet.
(3) I miss the lack of automation. With the last version I produced, I could click on one of the iterations, it would automatically be copied into the reference spot, the regeneration of the randomized iterations would happen, then the iteration array would automatically be copied into the clipboard. This saves a lot of steps in a very repetitive job. May be it is difficult to accomplish with the new designā¦ All I can say is that Iām getting repetitive strain injury
Suggested Enhancements:[/b]
I would like to see several PRESET buttons on the spreadsheet as follows:
(a) PRESET to equally distributed weighting. I generally start with equal distribution and work from there. I prefer to click on a button than enter the weights.
(b) PRESET to a rolling one pattern. The idea is to preset the iteration array so that one and only one node is enabled for each iteration.
(c) PRESET to a rolling zero pattern. The idea is to preset the iteration array so that one and only one node is disabled for each iteration. The remainder of the nodes have equally distributed weighting.
(d) PRESET to a randomized selection of two nodes - two nodes at random would be enabled for each iteration. The nodes would have equal weighting. This is for finding pairs of nodes that function well together.
(e) PRESET to a randomized array as per your separate spreadsheet.
Anyways, these are my thoughts. Excellent work Ralph.
Steve