 |
| Counterpropagation |
|
/******************************************************************************
==========================
Network: Counterpropagation Network
==========================
Application: Vision
Determination of the Angle of Rotation
Author: Karsten Kutza
Date: 30.4.96
Reference: R. Hecht-Nielsen
Counterpropagation Networks
Proceedings of the IEEE International Conference on
Neural Networks, II, IEEE Press, New York, NY, pp. 19-32,
1987
******************************************************************************/
/******************************************************************************
D E C L A R A T I O N S
******************************************************************************/
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
typedef int BOOL;
typedef char CHAR;
typedef int INT;
typedef double REAL;
#define FALSE 0
#define TRUE 1
#define NOT !
#define AND &&
#define OR ||
#define MIN_REAL -HUGE_VAL
#define MAX_REAL +HUGE_VAL
#define MIN(x,y) ((x)<(y) ? (x) : (y))
#define MAX(x,y) ((x)>(y) ? (x) : (y))
#define LO 0
#define HI 1
#define PI (2*asin(1))
#define sqr(x) ((x)*(x))
typedef struct { /* A LAYER OF A NET: */
INT Units; /* - number of units in this layer */
REAL* Output; /* - output of ith unit */
REAL** Weight; /* - connection weights to ith unit */
BOOL* Winner; /* - marker for winning instar */
} LAYER;
typedef struct { /* A NET: */
LAYER* InputLayer; /* - input layer */
LAYER* InstarLayer; /* - instar layer */
LAYER* OutstarLayer; /* - outstar layer */
INT Winners; /* - allowed number of winning instars */
REAL Alpha; /* - hidden layer learning rate */
REAL Beta; /* - output layer learning rate */
} NET;
/******************************************************************************
R A N D O M S D R A W N F R O M D I S T R I B U T I O N S
******************************************************************************/
void InitializeRandoms()
{
srand(4711);
}
INT RandomEqualINT(INT Low, INT High)
{
return rand() % (High-Low+1) + Low;
}
/******************************************************************************
A P P L I C A T I O N - S P E C I F I C C O D E
******************************************************************************/
#define NUM_DATA 8
#define X 11
#define Y 11
#define N (X * Y)
#define C NUM_DATA
#define M 2
CHAR Pattern[NUM_DATA][Y][X] = { { " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" OOOOO ",
" ",
" " },
{ " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
"OOOOO ",
" ",
" " },
{ " ",
" ",
" ",
" OO ",
" OOOOO ",
" OOOOOOO ",
" OOOOO ",
" OO ",
" ",
" ",
" " },
{ " ",
" ",
"OOOOO ",
" OOOOO ",
" OOO ",
" OOO ",
" OOO ",
" O ",
" O ",
" ",
" " },
{ " ",
" ",
" OOOOO ",
" OOOOO ",
" OOO ",
" OOO ",
" OOO ",
" O ",
" O ",
" ",
" " },
{ " ",
" ",
" OOOOO",
" OOOOO ",
" OOO ",
" OOO ",
" OOO ",
" O ",
" O ",
" ",
" " },
{ " ",
" ",
" ",
" OO ",
" OOOOO ",
" OOOOOOO ",
" OOOOO ",
" OO ",
" ",
" ",
" " },
{ " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" OOOOO",
" ",
" " } };
CHAR Pattern_[NUM_DATA][Y][X] = { { " ",
" ",
" O ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" ",
" " },
{ " ",
" ",
" O ",
" O ",
" O O ",
" O O ",
" O O ",
" O O ",
" O O ",
" ",
" " },
{ " ",
" ",
" ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" ",
" ",
" " },
{ " ",
" ",
" ",
" ",
" O ",
" O ",
" O ",
" OOO ",
" ",
" ",
" " },
{ " ",
" O ",
" O ",
" O ",
" OOO ",
" OO ",
" OOO O",
" OOOO ",
" OOOOO ",
" ",
" O " },
{ " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" OOOOO ",
" ",
" " },
{ " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
" OOOOO ",
" ",
" " },
{ " ",
" ",
" O ",
" O ",
" OOO ",
" OOO ",
" OOO ",
" OOOOO ",
"OOOOO ",
" ",
" " } };
REAL Input [NUM_DATA][N];
REAL Input_[NUM_DATA][N];
REAL Output[NUM_DATA][M];
FILE* f;
void NormalizeInput()
{
INT n,i;
REAL Length, Length_;
for (n=0; n<NUM_DATA; n++) {
Length = 0;
Length_ = 0;
for (i=0; i<N; i++) {
Length += sqr(Input [n][i]);
Length_ += sqr(Input_[n][i]);
}
Length = sqrt(Length);
Length_ = sqrt(Length_);
for (i=0; i<N; i++) {
Input [n][i] /= Length;
Input_[n][i] /= Length_;
}
}
}
void InitializeApplication(NET* Net)
{
INT n,i,j;
for (n=0; n<NUM_DATA; n++) {
for (i=0; i<Y; i++) {
for (j=0; j<X; j++) {
Input [n][i*X+j] = (Pattern [n][i][j] == 'O') ? HI : LO;
Input_[n][i*X+j] = (Pattern_[n][i][j] == 'O') ? HI : LO;
}
}
}
NormalizeInput();
for (n=0; n<NUM_DATA; n++) {
Output[n][0] = sin(n * 0.25 * PI);
Output[n][1] = cos(n * 0.25 * PI);
}
f = fopen("CPN.txt", "w");
}
void WriteInput(NET* Net, REAL* Input)
{
INT i;
for (i=0; i<N; i++) {
if (i%X == 0) {
fprintf(f, "\n");
}
fprintf(f, "%c", (Input[i] != LO) ? 'O' : ' ');
}
fprintf(f, " -> ");
}
void WriteOutput(NET* Net, REAL* Output)
{
REAL Angle;
Angle = (atan2(Output[0], Output[1]) / PI) * 180;
if (Angle < 0)
Angle = Angle + 360;
fprintf(f, "%0.0f°\n", Angle);
}
void FinalizeApplication(NET* Net)
{
fclose(f);
}
/******************************************************************************
I N I T I A L I Z A T I O N
******************************************************************************/
void GenerateNetwork(NET* Net)
{
INT i;
Net->InputLayer = (LAYER*) malloc(sizeof(LAYER));
Net->InstarLayer = (LAYER*) malloc(sizeof(LAYER));
Net->OutstarLayer = (LAYER*) malloc(sizeof(LAYER));
Net->InputLayer->Units = N;
Net->InputLayer->Output = (REAL*) calloc(N, sizeof(REAL));
Net->InstarLayer->Units = C;
Net->InstarLayer->Output = (REAL*) calloc(C, sizeof(REAL));
Net->InstarLayer->Weight = (REAL**) calloc(C, sizeof(REAL*));
Net->InstarLayer->Winner = (BOOL*) calloc(C, sizeof(BOOL));
Net->OutstarLayer->Units = M;
Net->OutstarLayer->Output = (REAL*) calloc(M, sizeof(REAL));
Net->OutstarLayer->Weight = (REAL**) calloc(M, sizeof(REAL*));
for (i=0; i<C; i++) {
Net->InstarLayer->Weight[i] = (REAL*) calloc(N, sizeof(REAL));
}
for (i=0; i<M; i++) {
Net->OutstarLayer->Weight[i] = (REAL*) calloc(C, sizeof(REAL));
}
Net->Winners = 1;
Net->Alpha = 0.1;
Net->Beta = 0.1;
}
void SetInput(NET* Net, REAL* Input, BOOL Protocoling)
{
INT i;
for (i=0; i<Net->InputLayer->Units; i++) {
Net->InputLayer->Output[i] = Input[i];
}
if (Protocoling) {
WriteInput(Net, Input);
}
}
void GetOutput(NET* Net, REAL* Output, BOOL Protocoling)
{
INT i;
for (i=0; i<Net->OutstarLayer->Units; i++) {
Output[i] = Net->OutstarLayer->Output[i];
}
if (Protocoling) {
WriteOutput(Net, Output);
}
}
/******************************************************************************
P R O P A G A T I N G S I G N A L S
******************************************************************************/
void PropagateToInstars(NET* Net)
{
INT w,i,j;
REAL Sum, SumWinners, MaxOut;
INT Winner;
for (i=0; i<Net->InstarLayer->Units; i++) {
Sum = 0;
for (j=0; j<Net->InputLayer->Units; j++) {
Sum += Net->InstarLayer->Weight[i][j] * Net->InputLayer->Output[j];
}
Net->InstarLayer->Output[i] = Sum;
Net->InstarLayer->Winner[i] = FALSE;
}
SumWinners = 0;
for (w=0; w<Net->Winners; w++) {
MaxOut = MIN_REAL;
for (i=0; i<Net->InstarLayer->Units; i++) {
if (NOT Net->InstarLayer->Winner[i] AND Net->InstarLayer->Output[i] > MaxOut)
MaxOut = Net->InstarLayer->Output[Winner = i];
}
Net->InstarLayer->Winner[Winner] = TRUE;
SumWinners += Net->InstarLayer->Output[Winner];
}
for (i=0; i<Net->InstarLayer->Units; i++) {
if (Net->InstarLayer->Winner[i])
Net->InstarLayer->Output[i] = Net->InstarLayer->Output[i] / SumWinners;
else
Net->InstarLayer->Output[i] = 0;
}
}
void PropagateToOutstars(NET* Net)
{
INT i,j;
REAL Sum;
for (i=0; i<Net->OutstarLayer->Units; i++) {
Sum = 0;
for (j=0; j<Net->InstarLayer->Units; j++) {
Sum += Net->OutstarLayer->Weight[i][j] * Net->InstarLayer->Output[j];
}
Net->OutstarLayer->Output[i] = Sum;
}
}
void PropagateNet(NET* Net)
{
PropagateToInstars(Net);
PropagateToOutstars(Net);
}
/******************************************************************************
T R A I N I N G T H E N E T
******************************************************************************/
INT Winner(NET* Net)
{
INT i;
for (i=0; i<Net->InstarLayer->Units; i++) {
if (Net->InstarLayer->Winner[i])
return i;
}
}
void TrainInstars(NET* Net, INT Epochs)
{
INT n,m,i,j;
for (i=0; i<Net->InstarLayer->Units; i++) {
for (j=0; j<Net->InputLayer->Units; j++) {
Net->InstarLayer->Weight[i][j] = Input[i][j];
}
}
Net->Winners = 1;
for (m=0; m<Epochs*NUM_DATA; m++) {
n = RandomEqualINT(0, NUM_DATA-1);
SetInput(Net, Input[n], FALSE);
PropagateToInstars(Net);
i = Winner(Net);
for (j=0; j<Net->InputLayer->Units; j++) {
Net->InstarLayer->Weight[i][j] +=
Net->Alpha * (Input[n][j] - Net->InstarLayer->Weight[i][j]);
}
}
}
void TrainOutstars(NET* Net, INT Epochs)
{
INT n,m,i,j;
for (i=0; i<Net->OutstarLayer->Units; i++) {
for (j=0; j<Net->InstarLayer->Units; j++) {
Net->OutstarLayer->Weight[i][j] = Output[j][i];
}
}
Net->Winners = 1;
for (m=0; m<Epochs*NUM_DATA; m++) {
n = RandomEqualINT(0, NUM_DATA-1);
SetInput(Net, Input[n], FALSE);
PropagateToInstars(Net);
j = Winner(Net);
for (i=0; i<Net->OutstarLayer->Units; i++) {
Net->OutstarLayer->Weight[i][j] +=
Net->Beta * (Output[n][i] - Net->OutstarLayer->Weight[i][j]);
}
}
}
/******************************************************************************
S I M U L A T I N G T H E N E T
******************************************************************************/
void SimulateNet(NET* Net, REAL* Input)
{
REAL Output[M];
SetInput(Net, Input, TRUE);
PropagateNet(Net);
GetOutput(Net, Output, TRUE);
}
/******************************************************************************
M A I N
******************************************************************************/
void main()
{
NET Net;
INT n;
InitializeRandoms();
GenerateNetwork(&Net);
InitializeApplication(&Net);
TrainInstars(&Net, 0); /* weights are computed for this application */
TrainOutstars(&Net, 0); /* weights are computed for this application */
Net.Winners = 2;
for (n=0; n<NUM_DATA; n++) {
SimulateNet(&Net, Input[n]);
}
for (n=0; n<NUM_DATA; n++) {
SimulateNet(&Net, Input_[n]);
}
FinalizeApplication(&Net);
}
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 0°
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 45°
OO
OOOOO
OOOOOOO
OOOOO
OO
-> 90°
OOOOO
OOOOO
OOO
OOO
OOO
O
O
-> 135°
OOOOO
OOOOO
OOO
OOO
OOO
O
O
-> 180°
OOOOO
OOOOO
OOO
OOO
OOO
O
O
-> 225°
OO
OOOOO
OOOOOOO
OOOOO
OO
-> 270°
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 315°
O
O
O
OOO
OOO
OOO
OOOOO
-> 0°
O
O
O O
O O
O O
O O
O O
-> 0°
O
OOO
OOO
OOO
OOOOO
-> 0°
O
O
O
OOO
-> 20°
O
O
O
OOO
OO
OOO O
OOOO
OOOOO
O -> 0°
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 0°
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 23°
O
O
OOO
OOO
OOO
OOOOO
OOOOO
-> 45°