initializegnn_static.m

介绍如何用较复杂的方法去优化对负载具有感知特性的“黑匣子”模型的神经网络控制。

function [W]=initializegnn_static(m,n,N_h,overlapfactor)
%[W]=initializegnn_static(m,n,N_h)
%this initializes a static Generalized Neural Network using something like 
%the Nguyen Widrow method [1]
%m=number of inputs (including a 1 for bias)
%n=number of outputs
%N_h=number of hidden neurons
%this is for bipolar activation functions
%
%Reference:
%Nguyen, D. and Widrow, B. "Improving the Learning Speed of 2-Layer
%Neural Networks By Choosing Initial Values for the Adaptive Weights", 
%Proceedings of the International Joint Conference on Neural Networks, San 
%Diego, pp. 21-26, 1990.

%    Copyright Travis Wiens 2008
%
%    This program is free software: you can redistribute it and/or modify
%    it under the terms of the GNU General Public License as published by
%    the Free Software Foundation, either version 3 of the License, or
%    (at your option) any later version.
%
%    This program is distributed in the hope that it will be useful,
%    but WITHOUT ANY WARRANTY; without even the implied warranty of
%    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
%    GNU General Public License for more details.
%
%    You should have received a copy of the GNU General Public License
%    along with this program.  If not, see <http://www.gnu.org/licenses/>.
%
%    If you would like to request a commerical (or other) license, please
%    feel free to contact [e-mail]

if nargin<4
	overlapfactor=0.7;%how much neurons overlap
end

W=zeros(n+m+N_h);

for i=1+m:(n+m+N_h)
    %use Nguyen Widrow
    N_inputs=i-2;%inputs to neuron (except 1)
    W_temp=2*rand(1,N_inputs)-1;%random direction (note this is biased away from axes)
    W_mag=overlapfactor*((n+m+N_h)/2)^(1/N_inputs);%magnitude of weight vector
    W_temp=W_mag/(sqrt(sum(W_temp.^2)))*W_temp;%scale direction to magnitude
    W(i,1:m-1)=W_temp(1:m-1);%skip over x=1 at m
    W(i,m+1:N_inputs+1)=W_temp(m:end);
    W(i,m)=(2*rand-1)*W_mag;
end