dear all morphometricians, at present I'm working on pattern recognition in a lepidopteran species-group. I would apply neural networks method (i.e., backpropagation networks) to relative warps of lepidopteran genitalia. but a collegue of mine from Mathematics dpt. told me that, as general rule, I could not apply nnet to negative values (!) To perform the analysis, am I allowed to modify all the value (x & y) of the relative warps, translating them along the Cartesian axes, but maintaining the reciprocal relationship among them, and convert all of them to positive values? and - more important - can someone tell me if really I couldn't apply nnet to negative values? Thank you very much in advance for any information, Angela Roggero -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
