Hey all,
Here are the minutes from lat meeting, but I wanted to say first that I am out of town this Friday and can't make it to the design meeting. Can we delay the next brain meeting to next Friday 10/20? Lemme know, thanks!
Minutes from 10/6
-Brain discussion
-Brain polynomial form:
w_i = ith synaptic weight
x_i = ith output
h_i = f(w_i,x_i) = SUM(sparse over n,m){ (a_nm) * (w_i)^n * (x_i)^m }
g(H) = SUM(over k) { (a_k)*PRODUCT(over i)[ (h_i)^(b_k) ] }
+ SUM(sparse over L) { (c_L) * SUM(over i) [ (h_i)^L ] }
g(H) acts on the set H of all h_i, treating each symmetrically. The output of g(H) is a scalar that is then run through an IIR filter. This output is then run through another polynomial d that acts like the output stage (sigmoid???).
-Discussion is needed concerning how chemical inputs are added.
-Energy costs also need to be attached to the brain. The current idea is to time the execution of the think function for the brain and use this to determine enrgy cost. Further debate may be needed. **This solution may not be consistent across platforms.**
-Synaptic weights are at the dendrites.
-Form needs to be specified for polynomials to determine other changeable brain variables such as synaptic weights should be plastic within a generation.
-Polynomials are fixed for a given generation and change only during evolution.
Cheers!
-Zach
---------------------------------
Talk is cheap. Use Yahoo! Messenger to make PC-to-Phone calls. Great rates starting at 1¢/min. |