Move static function prototypes from nbench1.h to appropriate files

git-svn-id: svn://mattst88.com/svn/cleanbench/trunk@14 0d43b9a7-5ab2-4d7b-af9d-f64450cef757

1 files changed, 70 insertions, 5 deletions

diff --git a/neural.c b/neural.c
index 04d49fe..a767fb4 100644
--- a/neural.c
+++ b/neural.c
@@ -6,11 +6,6 @@
 #include "nmglobal.h"
 #include "nbench1.h"
 
-/*
-** The Neural Net test requires an input data file.
-** The name is specified here.
-*/
-char *inpath="NNET.DAT";
 
 /********************************
 ** BACK PROPAGATION NEURAL NET **
@@ -37,6 +32,76 @@ char *inpath="NNET.DAT";
 ** on systems other than a Macintosh -- RG
 */
 
+/*
+** DEFINES
+*/
+#define T 1                     /* TRUE */
+#define F 0                     /* FALSE */
+#define ERR -1
+#define MAXPATS 10              /* max number of patterns in data file */
+#define IN_X_SIZE 5             /* number of neurodes/row of input layer */
+#define IN_Y_SIZE 7             /* number of neurodes/col of input layer */
+#define IN_SIZE 35              /* equals IN_X_SIZE*IN_Y_SIZE */
+#define MID_SIZE 8              /* number of neurodes in middle layer */
+#define OUT_SIZE 8              /* number of neurodes in output layer */
+#define MARGIN 0.1              /* how near to 1,0 do we have to come to stop? */
+#define BETA 0.09               /* beta learning constant */
+#define ALPHA 0.09              /* momentum term constant */
+#define STOP 0.1                /* when worst_error less than STOP, training is done */
+
+/*
+** The Neural Net test requires an input data file.
+** The name is specified here.
+*/
+char *inpath="NNET.DAT";
+
+/*
+** GLOBALS
+*/
+double  mid_wts[MID_SIZE][IN_SIZE];     /* middle layer weights */
+double  out_wts[OUT_SIZE][MID_SIZE];    /* output layer weights */
+double  mid_out[MID_SIZE];              /* middle layer output */
+double  out_out[OUT_SIZE];              /* output layer output */
+double  mid_error[MID_SIZE];            /* middle layer errors */
+double  out_error[OUT_SIZE];            /* output layer errors */
+double  mid_wt_change[MID_SIZE][IN_SIZE]; /* storage for last wt change */
+double  out_wt_change[OUT_SIZE][MID_SIZE]; /* storage for last wt change */
+double  in_pats[MAXPATS][IN_SIZE];      /* input patterns */
+double  out_pats[MAXPATS][OUT_SIZE];    /* desired output patterns */
+double  tot_out_error[MAXPATS];         /* measure of whether net is done */
+double  out_wt_cum_change[OUT_SIZE][MID_SIZE]; /* accumulated wt changes */
+double  mid_wt_cum_change[MID_SIZE][IN_SIZE];  /* accumulated wt changes */
+
+double  worst_error; /* worst error each pass through the data */
+double  average_error; /* average error each pass through the data */
+double  avg_out_error[MAXPATS]; /* average error each pattern */
+
+int iteration_count;    /* number of passes thru network so far */
+int numpats;            /* number of patterns in data file */
+int numpasses;          /* number of training passes through data file */
+int learned;            /* flag--if TRUE, network has learned all patterns */
+
+/*
+** PROTOTYPES
+*/
+static unsigned long DoNNetIteration(unsigned long nloops);
+static void do_mid_forward(int patt);
+static void do_out_forward();
+void display_output(int patt);
+static void do_forward_pass(int patt);
+static void do_out_error(int patt);
+static void worst_pass_error();
+static void do_mid_error();
+static void adjust_out_wts();
+static void adjust_mid_wts();
+static void do_back_pass(int patt);
+static void move_wt_changes();
+static int check_out_error();
+static void zero_changes();
+static void randomize_wts();
+static int read_data_file();
+/* static int initialize_net(); */
+
 /***********
 ** DoNNet **
 ************