Still holed up in my lab and finally got the artificial neural network to run and give me useful results. It actually did pretty well -- used a feedforward back-propagation flavored one after getting frustrated at trying to code perceptrons. It's a very trivial case: trying to predict productivity class based on 6 different Landsat bands + NDVI + EVI + elevation + aspect + slope, but at least it works. The real fun? Trying to map out the classification tree results because apparently CART doesn't give me outputs to stick back into ArcGIS like my neural network code that I did in MatLab... although CART did give me a nice set of tables and trees to stick in my write-up.
Darn if I got a penny for every profanity at the !#$& football fanatics uttered (actually uttered or thought of) today, I could probably afford a new Hoyt Ultra-Elite with a full set of Shrewd stabilizers, a CBE sight mount, a decent magnifying sight, a nice set of strings & cables from Winner's Choice, a matching T.R.U. Ball Sabertooth 360 4-finger release, and a dozen Carbon Express Nano arrows built and fletched to my specs. I could even here all the hubbub all the way from Newins-Ziegler and it's a difficult sprint from here to the football stadium.
Ok, think my lab computer has unfrozen up so I can resume working instead of thinking up what the Nine Layers of Hell would look like in my world.
No comments:
Post a Comment