double mfe_zscore(const char *seq, double mfe, int *type, int avoid_shuffle,
char* warning_string) {
double E, stdv, avg;
char *struc;
if (mfe>0) {
struc = space(strlen(seq)+1);
E = fold(seq, struc);
free(struc);
} else {
E=mfe;
}
avg = 0.0;
stdv = 0.0;
predict_values(seq, &avg, &stdv, type, avoid_shuffle, warning_string);
/* Just as backup strategy if something goes totally wrong,
we evaluate the sequence once again by shuffling */
if (avg > -1 || stdv < 0.1) {
if (*type == 2) *type = 3;
if (*type == 0) *type = 1;
predict_values(seq, &avg, &stdv, type, avoid_shuffle, warning_string);
}
/*printf("%f,%f\n",avg,stdv);*/
return ((E-avg)/stdv);
}
double LVlinear_predict_values(lvError *lvErr, const LVlinear_model *model_in, const LVArray_Hdl<LVlinear_node> x_in, LVArray_Hdl<double> dec_values_out){
try{
// Input validation: Uninitialized model
if (model_in == nullptr || model_in->w == nullptr || (*model_in->w)->dimSize == 0)
throw LVException(__FILE__, __LINE__, "Uninitialized model passed to liblinear_predict_values.");
// Input validation: Empty feature vector
if (x_in == nullptr || (*x_in)->dimSize == 0)
throw LVException(__FILE__, __LINE__, "Empty feature vector passed to liblinear_predict_values.");
// Input validation: Final index -1?
if ((*x_in)->elt[(*x_in)->dimSize - 1].index != -1)
throw LVException(__FILE__, __LINE__, "The index of the last element of the feature vector needs to be -1 (liblinear_predict_values).");
// Convert LVsvm_model to svm_model
auto mdl = std::make_unique<model>();
LVConvertModel(*model_in, *mdl);
int nr_class = model_in->nr_class;
int solver = (model_in->param).solver_type;
// Set up output array
int nr_dec = 0;
if (nr_class <= 2){
if (solver == MCSVM_CS)
nr_dec = 2;
else
nr_dec = 1;
}
else {
nr_dec = nr_class;
}
LVResizeNumericArrayHandle(dec_values_out, nr_dec);
(*dec_values_out)->dimSize = nr_dec;
double predicted_label = predict_values(mdl.get(), reinterpret_cast<feature_node*>((*x_in)->elt), (*dec_values_out)->elt);
return predicted_label;
}
catch (LVException &ex) {
ex.returnError(lvErr);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
catch (std::exception &ex) {
LVException::returnStdException(lvErr, __FILE__, __LINE__, ex);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
catch (...) {
LVException ex(__FILE__, __LINE__, "Unknown exception has occurred");
ex.returnError(lvErr);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
}
int copy_predict_values (char *predict, struct model *model_,
npy_intp *predict_dims, char *dec_values, int nr_class)
{
npy_intp i;
struct feature_node **predict_nodes;
predict_nodes = dense_to_sparse((double *) predict, predict_dims, model_->bias);
if (predict_nodes == NULL)
return -1;
for(i=0; i<predict_dims[0]; ++i) {
predict_values(model_, predict_nodes[i],
((double *) dec_values) + i*nr_class);
free(predict_nodes[i]);
}
free(predict_nodes);
return 0;
}
double LVlinear_predict_values(lvError *lvErr, const LVlinear_model *model_in, const LVArray_Hdl<LVlinear_node> x_in, LVArray_Hdl<double> dec_values_out){
try{
// Convert LVsvm_model to svm_model
std::unique_ptr<model> model(new model);
LVConvertModel(model_in, model.get());
int nr_class = model_in->nr_class;
int solver = (model_in->param).solver_type;
// Set up output array
int nr_dec = 0;
if (nr_class <= 2){
if (solver == MCSVM_CS)
nr_dec = 2;
else
nr_dec = 1;
}
else {
nr_dec = nr_class;
}
LVResizeNumericArrayHandle(dec_values_out, nr_dec);
(*dec_values_out)->dimSize = nr_dec;
double predicted_label = predict_values(model.get(), reinterpret_cast<feature_node*>((*x_in)->elt), (*dec_values_out)->elt);
return predicted_label;
}
catch (LVException &ex) {
ex.returnError(lvErr);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
catch (std::exception &ex) {
LVException::returnStdException(lvErr, __FILE__, __LINE__, ex);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
catch (...) {
LVException ex(__FILE__, __LINE__, "Unknown exception has occurred");
ex.returnError(lvErr);
(*dec_values_out)->dimSize = 0;
return std::nan("");
}
}
double SVMLinear::predictModel(vector<double> features)
{
if(modelLinearSVM==NULL)
{
fprintf(stdout,"Error, Train Model First \n");
return 0.0;
}
int nr_class=get_nr_class(modelLinearSVM);
int bias=modelLinearSVM->bias;
int sparsity=0.0;
for (int i=0; i<features.size(); i++)
if(features[i]!=0.0)
sparsity++;
feature_node *x=Malloc(struct feature_node,sparsity+bias+1); //bias and -1 index
int cnt=0;
for (int i=0; i<features.size(); i++)
{
if(features[i]!=0.0)
{
x[cnt].index=i+1;
x[cnt].value=features[i];
cnt++;
}
}
if(bias)
{
x[cnt].index=modelLinearSVM->nr_feature+1,
x[cnt].value=1;
cnt++;
}
x[cnt].index=-1;
double val=0;
predict_values(modelLinearSVM,x,&val);
return val;
}
int csr_copy_predict_values(npy_intp n_features, npy_intp *data_size,
char *data, npy_intp *index_size, char
*index, npy_intp *indptr_shape, char
*intptr, struct model *model_, char
*dec_values, int nr_class)
{
struct feature_node **predict_nodes;
npy_intp i;
predict_nodes = csr_to_sparse((double *) data, index_size,
(int *) index, indptr_shape, (int *) intptr, model_->bias, n_features);
if (predict_nodes == NULL)
return -1;
for (i = 0; i < indptr_shape[0] - 1; ++i) {
predict_values(model_, predict_nodes[i],
((double *) dec_values) + i*nr_class);
free(predict_nodes[i]);
}
free(predict_nodes);
return 0;
}
void ImageViewer::classify(QImage &img)
{
struct feature_node* x = Malloc(struct feature_node, NUM_FEATURES+1);
x[NUM_FEATURES].index = -1; // -1 marks the end of list
std::vector<double> vPredictValues;
double prob_estimates[1];
for (size_t i = 0; i < reader.getInstancesNumber(); i++) {
for (int j = 0; j < NUM_FEATURES; j++) {
x[j].index = 1+j; // 1-based feature number
x[j].value = reader.instancesFeatures[i][j];
}
int predict_label = predict(modelPedestrian, x);
if (predict_label == 1) {
predict_values(modelPedestrian, x, prob_estimates);
} else {
prob_estimates[0] = 0;
}
vPredictValues.push_back(prob_estimates[0]);
}
suppression(img, vPredictValues);
reader.instancesFeatures.clear();
free(x);
}
void do_predict(mxArray *plhs[], const mxArray *prhs[], struct model *model_, const int predict_probability_flag)
{
int label_vector_row_num, label_vector_col_num;
int feature_number, testing_instance_number;
int instance_index;
double *ptr_label, *ptr_predict_label;
double *ptr_prob_estimates, *ptr_dec_values, *ptr;
struct feature_node *x;
mxArray *pplhs[1]; // instance sparse matrix in row format
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_);
int nr_w;
double *prob_estimates=NULL;
if(nr_class==2 && model_->param.solver_type!=MCSVM_CS)
nr_w=1;
else
nr_w=nr_class;
// prhs[1] = testing instance matrix
feature_number = get_nr_feature(model_);
testing_instance_number = (int) mxGetM(prhs[1]);
if(col_format_flag)
{
feature_number = (int) mxGetM(prhs[1]);
testing_instance_number = (int) mxGetN(prhs[1]);
}
label_vector_row_num = (int) mxGetM(prhs[0]);
label_vector_col_num = (int) mxGetN(prhs[0]);
if(label_vector_row_num!=testing_instance_number)
{
mexPrintf("Length of label vector does not match # of instances.\n");
fake_answer(plhs);
return;
}
if(label_vector_col_num!=1)
{
mexPrintf("label (1st argument) should be a vector (# of column is 1).\n");
fake_answer(plhs);
return;
}
ptr_label = mxGetPr(prhs[0]);
// transpose instance matrix
if(col_format_flag)
pplhs[0] = (mxArray *)prhs[1];
else
{
mxArray *pprhs[1];
pprhs[0] = mxDuplicateArray(prhs[1]);
if(mexCallMATLAB(1, pplhs, 1, pprhs, "transpose"))
{
mexPrintf("Error: cannot transpose testing instance matrix\n");
fake_answer(plhs);
return;
}
}
prob_estimates = Malloc(double, nr_class);
plhs[0] = mxCreateDoubleMatrix(testing_instance_number, 1, mxREAL);
if(predict_probability_flag)
plhs[2] = mxCreateDoubleMatrix(testing_instance_number, nr_class, mxREAL);
else
plhs[2] = mxCreateDoubleMatrix(testing_instance_number, nr_w, mxREAL);
ptr_predict_label = mxGetPr(plhs[0]);
ptr_prob_estimates = mxGetPr(plhs[2]);
ptr_dec_values = mxGetPr(plhs[2]);
x = Malloc(struct feature_node, feature_number+2);
for(instance_index=0;instance_index<testing_instance_number;instance_index++)
{
int i;
double target_label, predict_label;
target_label = ptr_label[instance_index];
// prhs[1] and prhs[1]^T are sparse
read_sparse_instance(pplhs[0], instance_index, x, feature_number, model_->bias);
if(predict_probability_flag)
{
predict_label = predict_probability(model_, x, prob_estimates);
ptr_predict_label[instance_index] = predict_label;
for(i=0;i<nr_class;i++)
ptr_prob_estimates[instance_index + i * testing_instance_number] = prob_estimates[i];
}
else
{
double *dec_values = Malloc(double, nr_class);
predict_label = predict_values(model_, x, dec_values);
ptr_predict_label[instance_index] = predict_label;
for(i=0;i<nr_w;i++)
ptr_dec_values[instance_index + i * testing_instance_number] = dec_values[i];
free(dec_values);
}
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)
{
mexPrintf("Mean squared error = %g (regression)\n",error/total);
mexPrintf("Squared correlation coefficient = %g (regression)\n",
((total*sumpt-sump*sumt)*(total*sumpt-sump*sumt))/
((total*sumpp-sump*sump)*(total*sumtt-sumt*sumt))
);
}
//else
//mexPrintf("Accuracy = %g%% (%d/%d)\n", (double) correct/total*100,correct,total);
// return accuracy, mean squared error, squared correlation coefficient
plhs[1] = mxCreateDoubleMatrix(3, 1, mxREAL);
ptr = mxGetPr(plhs[1]);
ptr[0] = (double)correct/total*100;
ptr[1] = error/total;
ptr[2] = ((total*sumpt-sump*sumt)*(total*sumpt-sump*sumt))/
((total*sumpp-sump*sump)*(total*sumtt-sumt*sumt));
free(x);
if(prob_estimates != NULL)
free(prob_estimates);
}
void do_predict(mxArray *plhs[], const mxArray *prhs[], struct model *model_, const int predict_probability_flag)
{
int label_vector_row_num, label_vector_col_num;
int feature_number, testing_instance_number;
int instance_index;
double *ptr_instance, *ptr_label, *ptr_predict_label;
double *ptr_prob_estimates, *ptr_dec_values, *ptr;
struct feature_node *x;
mxArray *pplhs[1]; // instance sparse matrix in row format
int correct = 0;
int total = 0;
int nr_class=get_nr_class(model_);
int nr_classifier;
double *prob_estimates=NULL;
if(nr_class==2)
nr_classifier=1;
else
nr_classifier=nr_class;
// prhs[1] = testing instance matrix
feature_number = mxGetN(prhs[1]);
testing_instance_number = mxGetM(prhs[1]);
if(col_format_flag)
{
feature_number = mxGetM(prhs[1]);
testing_instance_number = mxGetN(prhs[1]);
}
label_vector_row_num = mxGetM(prhs[0]);
label_vector_col_num = mxGetN(prhs[0]);
if(label_vector_row_num!=testing_instance_number)
{
mexPrintf("Length of label vector does not match # of instances.\n");
fake_answer(plhs);
return;
}
if(label_vector_col_num!=1)
{
mexPrintf("label (1st argument) should be a vector (# of column is 1).\n");
fake_answer(plhs);
return;
}
ptr_instance = mxGetPr(prhs[1]);
ptr_label = mxGetPr(prhs[0]);
// transpose instance matrix
if(mxIsSparse(prhs[1]))
{
if(col_format_flag)
{
pplhs[0] = (mxArray *)prhs[1];
}
else
{
mxArray *pprhs[1];
pprhs[0] = mxDuplicateArray(prhs[1]);
if(mexCallMATLAB(1, pplhs, 1, pprhs, "transpose"))
{
mexPrintf("Error: cannot transpose testing instance matrix\n");
fake_answer(plhs);
return;
}
}
}
else
mexPrintf("Testing_instance_matrix must be sparse\n");
prob_estimates = Malloc(double, nr_class);
plhs[0] = mxCreateDoubleMatrix(testing_instance_number, 1, mxREAL);
if(predict_probability_flag)
plhs[2] = mxCreateDoubleMatrix(testing_instance_number, nr_class, mxREAL);
else
plhs[2] = mxCreateDoubleMatrix(testing_instance_number, nr_classifier, mxREAL);
ptr_predict_label = mxGetPr(plhs[0]);
ptr_prob_estimates = mxGetPr(plhs[2]);
ptr_dec_values = mxGetPr(plhs[2]);
x = Malloc(struct feature_node, feature_number+2);
for(instance_index=0;instance_index<testing_instance_number;instance_index++)
{
int i;
double target,v;
target = ptr_label[instance_index];
// prhs[1] and prhs[1]^T are sparse
read_sparse_instance(pplhs[0], instance_index, x, feature_number, model_->bias);
if(predict_probability_flag)
{
v = predict_probability(model_, x, prob_estimates);
ptr_predict_label[instance_index] = v;
for(i=0;i<nr_class;i++)
ptr_prob_estimates[instance_index + i * testing_instance_number] = prob_estimates[i];
}
else
{
double *dec_values = Malloc(double, nr_class);
v = predict(model_, x);
ptr_predict_label[instance_index] = v;
predict_values(model_, x, dec_values);
for(i=0;i<nr_classifier;i++)
ptr_dec_values[instance_index + i * testing_instance_number] = dec_values[i];
}
if(v == target)
++correct;
++total;
}
mexPrintf("Accuracy = %g%% (%d/%d)\n", (double)correct/total*100,correct,total);
// return accuracy, mean squared error, squared correlation coefficient
plhs[1] = mxCreateDoubleMatrix(1, 1, mxREAL);
ptr = mxGetPr(plhs[1]);
ptr[0] = (double)correct/total*100;
free(x);
if(prob_estimates != NULL)
free(prob_estimates);
}
double binary_class_cross_validation(const problem *prob, const parameter *param, int nr_fold)
{
dvec_t dec_values;
ivec_t ty;
int *labels;
if (nr_fold > 1)
{
int i;
int *fold_start = Malloc(int,nr_fold+1);
int l = prob->l;
int *perm = Malloc(int,l);
for(i=0;i<l;i++) perm[i]=i;
for(i=0;i<l;i++)
{
int j = i+rand()%(l-i);
std::swap(perm[i],perm[j]);
}
for(i=0;i<=nr_fold;i++)
fold_start[i]=i*l/nr_fold;
for(i=0;i<nr_fold;i++)
{
int begin = fold_start[i];
int end = fold_start[i+1];
int j,k;
struct problem subprob;
subprob.l = l-(end-begin);
subprob.x = Malloc(struct feature_node*,subprob.l);
subprob.y = Malloc(int,subprob.l);
k=0;
for(j=0;j<begin;j++)
{
subprob.x[k] = prob->x[perm[j]];
subprob.y[k] = prob->y[perm[j]];
++k;
}
for(j=end;j<l;j++)
{
subprob.x[k] = prob->x[perm[j]];
subprob.y[k] = prob->y[perm[j]];
++k;
}
struct model *submodel = train(&subprob,param);
//int svm_type = get_svm_type(submodel);
//if(svm_type == NU_SVR || svm_type == EPSILON_SVR){
// fprintf(stderr, "wrong svm type");
// exit(1);
//}
labels = Malloc(int, get_nr_class(submodel));
get_labels(submodel, labels);
if(get_nr_class(submodel) > 2)
{
fprintf(stderr,"Error: the number of class is not equal to 2\n");
exit(-1);
}
dec_values.resize(end);
ty.resize(end);
for(j=begin;j<end;j++) {
predict_values(submodel,prob->x[perm[j]], &dec_values[j]);
ty[j] = (prob->y[perm[j]] > 0)? 1: -1;
}
if(labels[0] <= 0) {
for(j=begin;j<end;j++)
dec_values[j] *= -1;
}
free_and_destroy_model(&submodel);
free(subprob.x);
free(subprob.y);
free(labels);
}
free(perm);
free(fold_start);
}
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);
}
void binary_class_predict(FILE *input, FILE *output){
int total = 0;
int *labels;
int max_nr_attr = 64;
struct feature_node *x = Malloc(struct feature_node, max_nr_attr);
dvec_t dec_values;
ivec_t true_labels;
int n;
if(model_->bias >= 1)
n = get_nr_feature(model_) + 1;
else
n = get_nr_feature(model_);
labels = Malloc(int, get_nr_class(model_));
get_labels(model_, 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 = -1; // strtol gives 0 if wrong format, and precomputed kernel has <index> start from 0
label = strtok(line," \t");
target_label = strtod(label,&endptr);
if(endptr == label)
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);
++i;
}
if(model_->bias >= 0){
x[i].index = n;
x[i].value = model_->bias;
++i;
}
x[i].index = -1;
predict_label = predict(model_,x);
fprintf(output,"%g\n",predict_label);
double dec_value;
predict_values(model_, x, &dec_value);
true_labels.push_back((target_label > 0)? 1: -1);
if(labels[0] <= 0) dec_value *= -1;
dec_values.push_back(dec_value);
}
validation_function(dec_values, true_labels);
free(labels);
free(x);
}