Talk:Neural Targeting

From Robowiki
Jump to navigation Jump to search

This page could definitely be fleshed out a bit with information about other neural targeting bots, but I just don't know enough about most of them to write intelligently about them. Some of them are: Orca, OrcaM, GB, TheBrainPi, NeuralPremier, Fe4r, Chomsky, Gaia, and Thinker. It also strikes me as reading a bit too historical and not quite techy enough, but really, neural targeting is a pretty broad field, so I think it might be OK to have it like this and link to various other technical pages / discussions. --Voidious 19:23, 12 November 2007 (UTC)

Both me and Wcsv used Self-Organized Maps (aka Kohonen Maps) as our Neural Network. I might release the code once I get it reformatted in a way that won't scare everyone off. --Chase-san 19:27, 12 November 2007 (UTC)

Can anyone told me briefly on how Neural Network and GHA algorithm work in simple (simpler than Wikipedia page) word? Tank you in advance. --Nat Pavasant 07:54, 1 November 2009 (UTC)

Interesting, I happened to start playing with Neural Networks yesterday. Currently I am experimenting with a 3-layer perceptron, which is a type of neural network (I believe). http://www.willamette.edu/~gorr/classes/cs449/figs/hidden1.gif is a simple diagram of a 3-layer perceptron. (Note that you could have more than 1 input; you could also have more than 1 hidden node, but for now we'll assume there is only one)

  1. Essentially, you feed it the input(s); each input, including the bias for the hidden node, is then multiplied by a weight and summed up.
  2. This is fed to the hidden node, which passes the sum through a function (typically something like tanh).
  3. This is then again, along with the bias for the output node, multiplied by their respective weights and summed up. That gives you your output

http://www.willamette.edu/~gorr/classes/cs449/intro.html has more information, specifically http://www.willamette.edu/~gorr/classes/cs449/multilayer.html for this 3-layer perceptron. As for training this...I'll have to explain that in another post. --Starrynte 15:47, 1 November 2009 (UTC)

Ahh neat, yeah, take a look at Kohonen Map! Cheers! --Chase 15:49, 1 November 2009 (UTC)

Thanks, even though I don't think I understand yet. Some more thing I want to know, what is the difference between back-propagation neural network (BPN) and feedforward neural network? --Nat Pavasant 12:44, 2 November 2009 (UTC)

Oh, just answer myself. back-propagation is training algorithm while feedforward is mapping (or whatever it call to get output) algorithm --Nat Pavasant 12:52, 2 November 2009 (UTC)
There are no threads on this page yet.