Commit Inicial

data/tanveer/waveform-noise/le_datos.m

@@ -0,0 +1,22 @@
+printf('lendo problema %s ...\n', problema);
+
+n_entradas= 40; n_clases= 3; n_fich= 1; fich{1}= 'waveform-+noise.data'; n_patrons(1)= 5000;
+
+n_max= max(n_patrons);
+x = zeros(n_fich, n_max, n_entradas); cl= zeros(n_fich, n_max);
+n_patrons_total = sum(n_patrons); n_iter=0;
+
+for i_fich=1:n_fich
+  f=fopen(fich{i_fich}, 'r');
+  if -1==f
+	error('erro en fopen abrindo %s\n', fich{i_fich});
+  end
+  for i=1:n_patrons(i_fich)
+  	fprintf(2,'%5.1f%%\r', 100*n_iter++/n_patrons_total);
+	for j = 1:n_entradas
+	  x(i_fich,i,j) = fscanf(f,'%g',1);
+	end	
+	cl(i_fich,i) = fscanf(f,'%i',1);  	% lectura da clase
+  end
+  fclose(f);
+end

data/tanveer/waveform-noise/waveform-+noise.c

@@ -0,0 +1,167 @@
+/* =====================================================================
+   David Aha
+   August 1988
+   Creates waveform domain data
+   Usage: create-waveform number num-attributes
+   See CART book, page 49 for details
+
+   Requires use of the UNIXSTAT tool named "probdist".
+   ===================================================================== */
+#include <stdio.h>
+#include <math.h>
+
+#define NUMBER_OF_ARGS 3
+#define NUMBER_OF_ATTRIBUTES 40
+#define NUMBER_OF_CLASSES 3
+
+/*==== Inputs ==== */
+int num_instances;
+unsigned seed;
+
+/*==== Main array of info ====*/
+float h[NUMBER_OF_CLASSES][NUMBER_OF_ATTRIBUTES];
+
+/* =====================================================================
+   Main Function
+   ===================================================================== */
+main(argc,argv)
+   int argc;
+   char *argv[];
+{
+   void execute();
+   void initialize();
+
+   if (argc != NUMBER_OF_ARGS) 
+     { printf("Usage: num-instances seed\n");
+       return(0);
+     }
+
+   num_instances = atoi(argv[1]);
+   seed = atoi(argv[2]);
+   initialize();
+   execute();
+}
+
+/* =====================================================================
+   Initializes the algorithm.
+   ===================================================================== */
+void initialize()
+{
+   void increment_seed();
+  
+   int i,j;
+   char command[100];
+
+   srandom(seed);
+
+   /*==== Setup for waveform of types 1 through 3 ====*/
+   for(i=0; i<3; i++)
+      for(j=0; j<40; j++)
+	h[i][j] = 0.0;
+
+   /*==== Waveform 1 ====*/
+   for(i=1; i<=6; i++)
+     h[0][i] = (float)i;
+   j=1;
+   for(i=11; i>=7; i--)
+     { h[0][i] = (float)j;
+       j++;
+     }
+
+   /*==== Waveform 2 ====*/
+   j = 1;
+   for(i=9; i<=14; i++)
+     { h[1][i] = (float)j;
+       j++;
+     }
+   j=1;
+   for(i=19; i>=15; i--)
+     { h[1][i] = (float)j;
+       j++;
+     }
+
+   /*==== Waveform 3 ====*/
+   j = 1;
+   for(i=5; i<=10; i++)
+     { h[2][i] = (float)j;
+       j++;
+     }
+   j=1;
+   for(i=15; i>=11; i--)
+     { h[2][i] = (float)j;
+       j++;
+     }
+
+   /*==== Create the required random values ====*/
+   system("/bin/rm -f temp");
+   sprintf(command,"probdist -s %d ran z %d > temp\n",
+	           seed,num_instances*NUMBER_OF_ATTRIBUTES);
+   system(command);
+   increment_seed();
+
+}
+   
+/* =====================================================================
+   Executes the algorithm.
+   ===================================================================== */
+void execute()
+{
+   void increment_seed();
+
+   int num_instance, num_attribute;
+   int waveform_type, choice[2];
+   FILE *fopen(), *fp;
+   float random_attribute_value, multiplier[2];
+   char line[100];
+   float val[40];
+
+   fp = fopen("temp","r");
+   for(num_instance=0; num_instance<num_instances; num_instance++)
+     {  /*==== Set up class type ====*/
+	waveform_type = (random() % 3);
+	switch (waveform_type)
+	  { case 0: choice[0] = 0; choice[1] = 1; break;
+	    case 1: choice[0] = 0; choice[1] = 2; break;
+	    case 2: choice[0] = 1; choice[1] = 2; break;
+	  }
+
+	/*==== Set up u and (1-u) for this call ====*/
+	multiplier[0] = (float)(random() % 1001)/1000.0;
+	multiplier[1] = 1.0 - multiplier[0];
+
+	/*==== Create the instance ====*/
+	for(num_attribute=0; num_attribute<NUMBER_OF_ATTRIBUTES;
+			     num_attribute++)
+	  { /*==== Get the base random attribute value ====*/
+	    fgets(line,100,fp);
+	    sscanf(line,"%f",&random_attribute_value);
+
+	    /*==== Calculate the value ====*/
+	    val[num_attribute] =
+	      (multiplier[0] * h[choice[0]][num_attribute]) +
+	      (multiplier[1] * h[choice[1]][num_attribute]) +
+	      random_attribute_value;
+	  }
+
+	/*==== Print the instance to the current output ====*/
+        printf("%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%4.2f,%d\n",
+	        val[0], val[1], val[2], val[3], val[4],
+		val[5], val[6], val[7], val[8], val[9],
+		val[10], val[11], val[12], val[13], val[14],
+		val[15], val[16], val[17], val[18], val[19], val[20],
+	        val[21], val[22], val[23], val[24],
+		val[25], val[26], val[27], val[28], val[29],
+		val[30], val[31], val[32], val[33], val[34],
+		val[35], val[36], val[37], val[38], val[39], waveform_type);
+      }
+   fclose(fp);
+   system("/bin/rm -f temp");
+}
+
+/* =====================================================================
+   Increments seed
+   ===================================================================== */
+void increment_seed()
+{
+   seed += 62010;
+}

data/tanveer/waveform-noise/waveform-+noise.names

@@ -0,0 +1,38 @@
+1. Title: Waveform Database Generator (written in C)
+ 
+2. Source:
+   (a) Breiman,L., Friedman,J.H., Olshen,R.A., & Stone,C.J. (1984). 
+       Classification and Regression Trees.  Wadsworth International
+       Group: Belmont, California.  (see pages 43-49).
+   (b) Donor: David Aha 
+   (c) Date: 11/10/1988
+
+3. Past Usage:
+     1. CART book (above):
+        -- Optimal Bayes classification rate: 86% accuracy
+        -- CART decision tree algorithm: 72%
+        -- Nearest Neighbor Algorithm: 38%
+           -- 300 training and 5000 test instances
+
+4. Relevant Information:
+     -- 3 classes of waves
+     -- 40 attributes, all of which include noise
+        -- The latter 19 attributes are all noise attributes with mean 0
+           and variance 1
+     -- See the book for details (49-55, 169)
+     -- waveform-+noise.data.Z contains 5000 instances
+
+5. Number of Instances: chosen by user
+
+6. Number of Attributes:
+    -- 40 attributes with continuous values between 0 and 6
+
+7. Attribute Information:
+    -- Each class is generated from a combination of 2 of 3 "base" waves
+    -- Each instance is generated f added noise (mean 0, variance 1) in 
+       each attribute
+     -- See the book for details (49-55, 169)
+    
+8. Missing Attribute Values: none
+
+9. Class Distribution: 33% for each of 3 classes

data/tanveer/waveform-noise/waveform-noise.cost

@@ -0,0 +1,6 @@
+% Rows	Columns
+3	3
+% Matrix elements
+ 0.0	 1.0	 1.0	
+ 1.0	 0.0	 1.0	
+ 1.0	 1.0	 0.0	

data/tanveer/waveform-noise/waveform-noise.txt

@@ -0,0 +1,8 @@
+n_entradas= 40
+n_clases= 3
+n_arquivos= 1
+fich1= waveform-noise_R.dat
+n_patrons1= 5000
+n_patrons_entrena= 2500
+n_patrons_valida= 2500
+n_conxuntos= 1