Class J's - trackworthiness?
nw-modeling-list at nwhs.org
nw-modeling-list at nwhs.org
Thu Oct 5 16:59:25 EDT 2006
Hi Gary,
I love seeing a long post about an engineering topic--figure it must be one of yours and I'm sure to learn a lot. Would you mind if I quibble just a bit about one statement? You wrote: "Even with computers, only an iterative solution is possible (iterative = run the numbers and see what comes out)." Iterative in this sense refers to a mathematical method based on the calculus, such as Newton-Raphson, that starts with empirical data involving measurements on several variables and a model equation based on physical theory that interrelates those variables through a set of parameters whose ideal values are unknown; "ideal" being defined as those values that minimize some other "loss" or "objective" function. If the model equation is linear in the parameters, then generally an iterative solution is not required, as in multivariate linear regression. Here calculus isn't involved, only algebra. Iteration involves finding an approximate solution, then using that solution as a starting
point to find a better solution, and continuing this process until the gain as measured by the decrease in the loss function is negligible.
I must be getting old. My pedantry on the point reminds me of a manager at Illinois Bell from thirty years ago, long before email. About 10-15 paper memos a day would circulate through the department, with a little sheet stapled to each where you'd check off your initials after reading it. Frank C. would go through every one and mark it up with spelling, punctuation and grammar corrections before passing it along.
Sorry guys.
Tom Leuthner
P.S. I believe someone on the list modeling the N&W lives in the greater Chicago area. Ever have an open house? What scale?
More information about the NW-Modeling-List
mailing list