Mercurial > hg > camir-aes2014

comparison toolboxes/SVM-light/src/svm_classify.c @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 /***********************************************************************/
2 /* */
3 /* svm_classify.c */
4 /* */
5 /* Classification module of Support Vector Machine. */
6 /* */
7 /* Author: Thorsten Joachims */
8 /* Date: 02.07.02 */
9 /* */
10 /* Copyright (c) 2002 Thorsten Joachims - All rights reserved */
11 /* */
12 /* This software is available for non-commercial use only. It must */
13 /* not be modified and distributed without prior permission of the */
14 /* author. The author is not responsible for implications from the */
15 /* use of this software. */
16 /* */
17 /************************************************************************/
18
19 # include "svm_common.h"
20
21 char docfile[200];
22 char modelfile[200];
23 char predictionsfile[200];
24
25 void read_input_parameters(int, char **, char *, char *, char *, long *,
26 long *);
27 void print_help(void);
28
29
30 int main (int argc, char* argv[])
31 {
32 DOC *doc; /* test example */
33 WORD *words;
34 long max_docs,max_words_doc,lld;
35 long totdoc=0,queryid,slackid;
36 long correct=0,incorrect=0,no_accuracy=0;
37 long res_a=0,res_b=0,res_c=0,res_d=0,wnum,pred_format;
38 long j;
39 double t1,runtime=0;
40 double dist,doc_label,costfactor;
41 char *line,*comment;
42 FILE *predfl,*docfl;
43 MODEL *model;
44
45 read_input_parameters(argc,argv,docfile,modelfile,predictionsfile,
46 &verbosity,&pred_format);
47
48 nol_ll(docfile,&max_docs,&max_words_doc,&lld); /* scan size of input file */
49 max_words_doc+=2;
50 lld+=2;
51
52 line = (char *)my_malloc(sizeof(char)*lld);
53 words = (WORD *)my_malloc(sizeof(WORD)*(max_words_doc+10));
54
55 model=read_model(modelfile);
56
57 if(model->kernel_parm.kernel_type == 0) { /* linear kernel */
58 /* compute weight vector */
59 add_weight_vector_to_linear_model(model);
60 }
61
62 if(verbosity>=2) {
63 printf("Classifying test examples.."); fflush(stdout);
64 }
65
66 if ((docfl = fopen (docfile, "r")) == NULL)
67 { perror (docfile); exit (1); }
68 if ((predfl = fopen (predictionsfile, "w")) == NULL)
69 { perror (predictionsfile); exit (1); }
70
71 while((!feof(docfl)) && fgets(line,(int)lld,docfl)) {
72 if(line[0] == '#') continue; /* line contains comments */
73 parse_document(line,words,&doc_label,&queryid,&slackid,&costfactor,&wnum,
74 max_words_doc,&comment);
75 totdoc++;
76 if(model->kernel_parm.kernel_type == 0) { /* linear kernel */
77 for(j=0;(words[j]).wnum != 0;j++) { /* Check if feature numbers */
78 if((words[j]).wnum>model->totwords) /* are not larger than in */
79 (words[j]).wnum=0; /* model. Remove feature if */
80 } /* necessary. */
81 doc = create_example(-1,0,0,0.0,create_svector(words,comment,1.0));
82 t1=get_runtime();
83 dist=classify_example_linear(model,doc);
84 runtime+=(get_runtime()-t1);
85 free_example(doc,1);
86 }
87 else { /* non-linear kernel */
88 doc = create_example(-1,0,0,0.0,create_svector(words,comment,1.0));
89 t1=get_runtime();
90 dist=classify_example(model,doc);
91 runtime+=(get_runtime()-t1);
92 free_example(doc,1);
93 }
94 if(dist>0) {
95 if(pred_format==0) { /* old weired output format */
96 fprintf(predfl,"%.8g:+1 %.8g:-1\n",dist,-dist);
97 }
98 if(doc_label>0) correct++; else incorrect++;
99 if(doc_label>0) res_a++; else res_b++;
100 }
101 else {
102 if(pred_format==0) { /* old weired output format */
103 fprintf(predfl,"%.8g:-1 %.8g:+1\n",-dist,dist);
104 }
105 if(doc_label<0) correct++; else incorrect++;
106 if(doc_label>0) res_c++; else res_d++;
107 }
108 if(pred_format==1) { /* output the value of decision function */
109 fprintf(predfl,"%.8g\n",dist);
110 }
111 if((int)(0.01+(doc_label*doc_label)) != 1)
112 { no_accuracy=1; } /* test data is not binary labeled */
113 if(verbosity>=2) {
114 if(totdoc % 100 == 0) {
115 printf("%ld..",totdoc); fflush(stdout);
116 }
117 }
118 }
119 free(line);
120 free(words);
121 free_model(model,1);
122
123 if(verbosity>=2) {
124 printf("done\n");
125
126 /* Note by Gary Boone Date: 29 April 2000 */
127 /* o Timing is inaccurate. The timer has 0.01 second resolution. */
128 /* Because classification of a single vector takes less than */
129 /* 0.01 secs, the timer was underflowing. */
130 printf("Runtime (without IO) in cpu-seconds: %.2f\n",
131 (float)(runtime/100.0));
132
133 }
134 if((!no_accuracy) && (verbosity>=1)) {
135 printf("Accuracy on test set: %.2f%% (%ld correct, %ld incorrect, %ld total)\n",(float)(correct)*100.0/totdoc,correct,incorrect,totdoc);
136 printf("Precision/recall on test set: %.2f%%/%.2f%%\n",(float)(res_a)*100.0/(res_a+res_b),(float)(res_a)*100.0/(res_a+res_c));
137 }
138
139 return(0);
140 }
141
142 void read_input_parameters(int argc, char **argv, char *docfile,
143 char *modelfile, char *predictionsfile,
144 long int *verbosity, long int *pred_format)
145 {
146 long i;
147
148 /* set default */
149 strcpy (modelfile, "svm_model");
150 strcpy (predictionsfile, "svm_predictions");
151 (*verbosity)=2;
152 (*pred_format)=1;
153
154 for(i=1;(i<argc) && ((argv[i])[0] == '-');i++) {
155 switch ((argv[i])[1])
156 {
157 case 'h': print_help(); exit(0);
158 case 'v': i++; (*verbosity)=atol(argv[i]); break;
159 case 'f': i++; (*pred_format)=atol(argv[i]); break;
160 default: printf("\nUnrecognized option %s!\n\n",argv[i]);
161 print_help();
162 exit(0);
163 }
164 }
165 if((i+1)>=argc) {
166 printf("\nNot enough input parameters!\n\n");
167 print_help();
168 exit(0);
169 }
170 strcpy (docfile, argv[i]);
171 strcpy (modelfile, argv[i+1]);
172 if((i+2)<argc) {
173 strcpy (predictionsfile, argv[i+2]);
174 }
175 if(((*pred_format) != 0) && ((*pred_format) != 1)) {
176 printf("\nOutput format can only take the values 0 or 1!\n\n");
177 print_help();
178 exit(0);
179 }
180 }
181
182 void print_help(void)
183 {
184 printf("\nSVM-light %s: Support Vector Machine, classification module %s\n",VERSION,VERSION_DATE);
185 copyright_notice();
186 printf(" usage: svm_classify [options] example_file model_file output_file\n\n");
187 printf("options: -h -> this help\n");
188 printf(" -v [0..3] -> verbosity level (default 2)\n");
189 printf(" -f [0,1] -> 0: old output format of V1.0\n");
190 printf(" -> 1: output the value of decision function (default)\n\n");
191 }
192
193
194
195