C++ Neural Networks and Fuzzy Logic:The Kohonen Self-Organizing Map
Click Here!
function GetCookie (name)
{
var arg = name + "=";
var alen = arg.length;
var clen = document.cookie.length;
var i = 0;
while (i < clen)
{
var j = i + alen;
if (document.cookie.substring(i, j) == arg) {
var end = document.cookie.indexOf (";", j);
if (end == -1)
end = document.cookie.length;
return unescape(document.cookie.substring(j, end));
}
i = document.cookie.indexOf(" ", i) + 1;
if (i == 0) break;
}
return null;
}
var m1='';
var gifstr=GetCookie("UsrType");
if((gifstr!=0 ) && (gifstr!=null)) { m2=gifstr; }
document.write(m1+m2+m3);
Keyword
Title
Author
ISBN
Publisher
Imprint
Brief
Full
Advanced Search
Search Tips
Please Select
-----------
Components
Content Mgt
Certification
Databases
Enterprise Mgt
Fun/Games
Groupware
Hardware
Intranet Dev
Middleware
Multimedia
Networks
OS
Prod Apps
Programming
Security
UI
Web Services
Webmaster
Y2K
-----------
New Titles
-----------
Free Archive
To access the contents, click the chapter and section titles.
C++ Neural Networks and Fuzzy Logic
(Publisher: IDG Books Worldwide, Inc.)
Author(s): Valluru B. Rao
ISBN: 1558515526
Publication Date: 06/01/95
Search this book:
Previous Table of Contents Next
Kohonen_layer:
• float * weights Pointer to the weights matrix. • int winner_index Index value of the output, which is the winner. • float win_distance The Euclidean distance of the winner weight vector from the input vector. • int neighborhood_size The size of the neighborhood. • Kohonen_layer(int, int, int) Constructor for the layer: inputs, outputs, and the neighborhood size. • ~Kohonen_layer() Destructor. • virtual void calc_out() The function to calculate the outputs; for the Kohonen layer this models lateral competition. • void randomize_weights() A function to initialize weights with random normal values. • void update_neigh_size(nt) This function updates the neighborhood size with a new value. • void update_weights(const float) This function updates the weights according to the training law using the passed parameter, alpha. • void list_weights() This function can be used to list the weight matrix. • void list_outputs() This function is used to write outputs to the output file. • float get_win_dist() Returns the Euclidean distance between the winner weight vector and the input vector.
Kohonen_network:
• layer *layer_ptr[2] Pointer array; element 0 points to the input layer, element 1 points to the Kohonen layer. • int layer_size[2] Array of layer sizes for the two layers. • int neighborhood_size The current neighborhood size. • Kohonen_network() Constructor. • ~Kohonen_network() Destructor. • void get_layer_info() Gets information about the layer sizes. • void set_up_network(int) Connects layers and sets up the Kohonen map. • void randomize_weights() Creates random normalized weights. • void update_neigh_size(int) Changes the neighborhood size. • void update_weights(const float) Performs weight update according to the training law. • void list_weights() Can be used to list the weight matrix. • void list_outputs() Can be used to list outputs. • void get_next_vector(FILE *) Function gets another input vector from the input file. • void process_next_pattern() Applies pattern to the Kohonen map. • float get_win_dist() Returns the winner’s distance from the input vector. • int get_win_index() Returns the index of the winner.
Implementation of the Kohonen Layer and Kohonen Network Listing 11.3 shows the layerk.cpp file, which has the implementation of the functions outlined. Listing 11.3 Implementation file for the Kohonen layer and Kohonen network :layerk.cpp
// layerk.cpp V.Rao, H.Rao // compile for floating point hardware if available #include “layer.cpp” #include “layerk.h”
Kohonen_layer::Kohonen_layer(int i, int o, int init_neigh_size) { num_inputs=i; num_outputs=o; neighborhood_size=init_neigh_size; weights = new float[num_inputs*num_outputs]; outputs = new float[num_outputs]; }
Kohonen_layer::~Kohonen_layer() { delete [num_outputs*num_inputs] weights; delete [num_outputs] outputs; } void Kohonen_layer::calc_out() { // implement lateral competition // choose the output with the largest // value as the winner; neighboring // outputs participate in next weight // update. Winner’s output is 1 while // all other outputs are zero
int i,j,k; float accumulator=0.0; float maxval; winner_index=0; maxval=-1000000;
for (j=0; j<num_outputs; j++) {
for (i=0; i<num_inputs; i++)
{ k=i*num_outputs; if (weights[k+j]*weights[k+j] > 1000000.0) { cout << “weights are blowing up\n”; cout << “try a smaller learning constant\n”; cout << “e.g. beta=0.02 aborting...\n”; exit(1); } outputs[j]=weights[k+j]*(*(inputs+i)); accumulator+=outputs[j]; } // no squash function outputs[j]=accumulator; if (outputs[j] > maxval) { maxval=outputs[j]; winner_index=j; } accumulator=0; }
// set winner output to 1 outputs[winner_index]=1.0; // now zero out all other outputs for (j=0; j< winner_index; j++) outputs[j]=0; for (j=num_outputs-1; j>winner_index; j—) outputs[j]=0;
}
void Kohonen_layer::randomize_weights() { int i, j, k; const unsigned first_time=1;
void Kohonen_layer::update_weights(const float alpha) { int i, j, k; int start_index, stop_index; // learning law: weight_change = // alpha*(input-weight) // zero change if input and weight // vectors are aligned // only update those outputs that // are within a neighborhood’s distance // from the last winner start_index = winner_index - neighborhood_size;
if (start_index < 0) start_index =0;
stop_index = winner_index + neighborhood_size;
if (stop_index > num_outputs-1) stop_index = num_outputs-1;
for (i=0; i< num_inputs; i++) { k=i*num_outputs; for (j=start_index; j<=stop_index; j++) weights[k+j] += alpha*((*(inputs+i))-weights[k+j]); }
}
void Kohonen_layer::list_weights() { int i, j, k;
for (i=0; i< num_inputs; i++) { k=i*num_outputs; for (j=0; j< num_outputs; j++) cout << “weight[“<<i<<”,”<< j<<”] is: “<<weights[k+j]; } }
void Kohonen_layer::list_outputs() { int i;
for (i=0; i< num_outputs; i++) { cout << “outputs[“<<i<< “] is: “<<outputs[i]; }
}
float Kohonen_layer::get_win_dist() { int i, j, k; j=winner_index; float accumulator=0;
float * win_dist_vec = new float [num_inputs];
for (i=0; i< num_inputs; i++) { k=i*num_outputs; win_dist_vec[i]=(*(inputs+i))-weights[k+j]; accumulator+=win_dist_vec[i]*win_dist_vec[i]; }
win_distance =(float)sqrt((double)accumulator);
delete [num_inputs]win_dist_vec;
return win_distance;
}
Kohonen_network::Kohonen_network() {
}
Kohonen_network::~Kohonen_network() { }
void Kohonen_network::get_layer_info() { int i;
//————————————————————— // // Get layer sizes for the Kohonen network // // ————————————————————-
cout << “ Enter in the layer sizes separated by spaces.\n”; cout << “ A Kohonen network has an input layer \n”; cout << “ followed by a Kohonen (output) layer \n”;
} void Kohonen_network::set_up_network(int nsz) { int i;
// set up neighborhood size neighborhood_size = nsz;
//———————————————————————————- // Construct the layers // //———————————————————————————-
layer_ptr[0] = new input_layer(0,layer_size[0]);
layer_ptr[1] = new Kohonen_layer(layer_size[0], layer_size[1],neighborhood_size);
for (i=0;i<2;i++) { if (layer_ptr[i] == 0) { cout << “insufficient memory\n”; cout << “use a smaller architecture\n”; exit(1); } }
//———————————————————————————- // Connect the layers // //———————————————————————————- // set inputs to previous layer outputs for the Kohonen layer layer_ptr[1]->inputs = layer_ptr[0]->outputs;