// Revised from function 'svm_predict_probability' for HIK
double hik_predict_probability(const svm_model *model,Array2dC<double>& eval,const int m,const int upper_bound,svm_inode *x,double *prob_estimates)
{
if ((model->param.svm_type == C_SVC || model->param.svm_type == NU_SVC) &&
model->probA!=NULL && model->probB!=NULL)
{
int i;
int nr_class = model->nr_class;
double *dec_values = new double[nr_class*(nr_class-1)/2];
hik_predict_values(model, eval, m, upper_bound, x, dec_values);
double min_prob=1e-7;
double **pairwise_prob=new double*[nr_class];
for(i=0;i<nr_class;i++)
pairwise_prob[i]=new double[nr_class];
int k=0;
for(i=0;i<nr_class;i++)
{
for(int j=i+1;j<nr_class;j++)
{
pairwise_prob[i][j]=std::min(std::max(sigmoid_predict(dec_values[k],model->probA[k],model->probB[k]),min_prob),1-min_prob);
pairwise_prob[j][i]=1-pairwise_prob[i][j];
k++;
}
}
multiclass_probability(nr_class,pairwise_prob,prob_estimates);
int prob_max_idx = 0;
for(i=1;i<nr_class;i++)
if(prob_estimates[i] > prob_estimates[prob_max_idx])
prob_max_idx = i;
for(i=0;i<nr_class;i++) delete[] pairwise_prob[i];
delete[] pairwise_prob;
delete[] dec_values;
return model->label[prob_max_idx];
}
else
return hik_predict(model, eval, m, upper_bound, x);
}
void do_predict(FILE *input, FILE *output, FILE *output2)
{
int correct = 0;
int total = 0;
double error = 0;
double sump = 0, sumt = 0, sumpp = 0, sumtt = 0, sumpt = 0;
int nr_class=get_nr_class(model_);
double *dec_values = (double *) malloc(nr_class*sizeof(double));
double prob = 0;
double *prob_estimates=NULL;
int j, n;
int nr_feature=get_nr_feature(model_);
if(model_->bias>=0)
n=nr_feature+1;
else
n=nr_feature;
if(flag_predict_probability)
{
int *labels;
if(!check_probability_model(model_))
{
fprintf(stderr, "probability output is only supported for logistic regression\n");
exit(1);
}
labels=(int *) malloc(nr_class*sizeof(int));
get_labels(model_,labels);
prob_estimates = (double *) malloc(nr_class*sizeof(double));
fprintf(output,"labels");
for(j=0;j<nr_class;j++)
fprintf(output," %d",labels[j]);
fprintf(output,"\n");
free(labels);
}
max_line_len = 1024;
line = (char *)malloc(max_line_len*sizeof(char));
while(readline(input) != NULL)
{
int i = 0;
double target_label, predict_label;
char *idx, *val, *label, *endptr;
int inst_max_index = 0; // strtol gives 0 if wrong format
label = strtok(line," \t\n");
if(label == NULL) // empty line
exit_input_error(total+1);
target_label = strtod(label,&endptr);
if(endptr == label || *endptr != '\0')
exit_input_error(total+1);
while(1)
{
if(i>=max_nr_attr-2) // need one more for index = -1
{
max_nr_attr *= 2;
x = (struct feature_node *) realloc(x,max_nr_attr*sizeof(struct feature_node));
}
idx = strtok(NULL,":");
val = strtok(NULL," \t");
if(val == NULL)
break;
errno = 0;
x[i].index = (int) strtol(idx,&endptr,10);
if(endptr == idx || errno != 0 || *endptr != '\0' || x[i].index <= inst_max_index)
exit_input_error(total+1);
else
inst_max_index = x[i].index;
errno = 0;
x[i].value = strtod(val,&endptr);
if(endptr == val || errno != 0 || (*endptr != '\0' && !isspace(*endptr)))
exit_input_error(total+1);
// feature indices larger than those in training are not used
if(x[i].index <= nr_feature)
++i;
}
if(model_->bias>=0)
{
x[i].index = n;
x[i].value = model_->bias;
i++;
}
x[i].index = -1;
if(flag_predict_probability)
{
int j;
predict_label = predict_probability(model_,x,prob_estimates);
fprintf(output,"%g",predict_label);
for(j=0;j<model_->nr_class;j++)
fprintf(output," %g",prob_estimates[j]);
fprintf(output,"\n");
}
else
{
predict_label=predict_values(model_, x, dec_values);
fprintf(output,"%g\n",predict_label);
//transfer the score into probability
prob = sigmoid_predict(dec_values[0], model_->probA, model_->probB);
fprintf(output2,"%4g\n",prob);
}
if(predict_label == target_label)
++correct;
error += (predict_label-target_label)*(predict_label-target_label);
sump += predict_label;
sumt += target_label;
sumpp += predict_label*predict_label;
sumtt += target_label*target_label;
sumpt += predict_label*target_label;
++total;
}
if(model_->param.solver_type==L2R_L2LOSS_SVR ||
model_->param.solver_type==L2R_L1LOSS_SVR_DUAL ||
model_->param.solver_type==L2R_L2LOSS_SVR_DUAL)
{
info("Mean squared error = %g (regression)\n",error/total);
info("Squared correlation coefficient = %g (regression)\n",
((total*sumpt-sump*sumt)*(total*sumpt-sump*sumt))/
((total*sumpp-sump*sump)*(total*sumtt-sumt*sumt))
);
}
else
info("Accuracy = %g%% (%d/%d)\n",(double) correct/total*100,correct,total);
if(flag_predict_probability)
free(prob_estimates);
free(dec_values);
}