SVECTOR *psi(PATTERNX x, LABEL y, STRUCTMODEL *sm,
STRUCT_LEARN_PARM *sparm)
{
/* Returns a feature vector describing the match between pattern x and
label y. The feature vector is returned as an SVECTOR
(i.e. pairs <featurenumber:featurevalue>), where the last pair has
featurenumber 0 as a terminator. Featurenumbers start with 1 and end with
sizePsi. This feature vector determines the linear evaluation
function that is used to score labels. There will be one weight in
sm.w for each feature. Note that psi has to match
find_most_violated_constraint_???(x, y, sm) and vice versa. In
particular, find_most_violated_constraint_???(x, y, sm) finds that
ybar!=y that maximizes psi(x,ybar,sm)*sm.w (where * is the inner
vector product) and the appropriate function of the loss. */
SVECTOR *fvec;
if (sparm->loss_function == 2) {
/* specific psi for bilinear SVM optimization */
fvec = smult_s(x.doc->fvec, y.classlabel==2 ? 0.5 : -0.5);
fvec->words[0].weight = 0.0; /* nullify 'theta' */
} else {
/* shift the feature numbers to the position of weight vector of class y */
fvec = shift_s(x.doc->fvec, (y.classlabel-1)*sparm->num_features);
}
/* The following makes sure that the weight vectors for each class
are treated separately when kernels are used . */
fvec->kernel_id = y.classlabel;
return(fvec);
}
SVECTOR* add_list_ss(SVECTOR *a)
/* computes the linear combination of the SVECTOR list weighted
by the factor of each SVECTOR */
{
SVECTOR *scaled,*oldsum,*sum,*f;
WORD empty[2];
if(a){
sum=smult_s(a,a->factor);
for(f=a->next;f;f=f->next) {
scaled=smult_s(f,f->factor);
oldsum=sum;
sum=add_ss(sum,scaled);
free_svector(oldsum);
free_svector(scaled);
}
sum->factor=1.0;
}
else {
empty[0].wnum=0;
sum=create_svector(empty,"",1.0);
}
return(sum);
}
SVECTOR* add_list_ss_r(SVECTOR *a, double min_non_zero)
/* computes the linear combination of the SVECTOR list weighted
by the factor of each SVECTOR */
{
SVECTOR *oldsum,*sum,*f;
WORD empty[2];
if(!a) {
empty[0].wnum=0;
sum=create_svector(empty,NULL,1.0);
}
else if(a && (!a->next)) {
sum=smult_s(a,a->factor);
}
else {
sum=multadd_ss_r(a,a->next,a->factor,a->next->factor,min_non_zero);
for(f=a->next->next;f;f=f->next) {
oldsum=sum;
sum=multadd_ss_r(oldsum,f,1.0,f->factor,min_non_zero);
free_svector(oldsum);
}
}
return(sum);
}
SVECTOR *psi(PATTERN x, LABEL y, STRUCTMODEL *sm, STRUCT_LEARN_PARM *sparm) {
/*
Creates the feature vector \Psi(x,y) and return a pointer to
sparse vector SVECTOR in SVM^light format. The dimension of the
feature vector returned has to agree with the dimension in sm->sizePsi.
*/
SVECTOR *fvec=NULL;
SVECTOR *psi1=NULL;
SVECTOR *psi2=NULL;
SVECTOR *temp_psi=NULL;
SVECTOR *temp_sub=NULL;
WORD *words = NULL;
words = (WORD *) malloc(sizeof(WORD));
if(!words) die("Memory error.");
words[0].wnum = 0;
words[0].weight = 0;
fvec = create_svector(words,"",1);
psi1 = create_svector(words,"",1);
psi2 = create_svector(words,"",1);
free(words);
int i,j = 0;
for (i = 0; i < (x.n_pos+x.n_neg); i++){
if(y.labels[i] == 1){
temp_psi = add_ss(psi1, x.x_is[i].phi1phi2_pos);
}
else{
temp_psi = add_ss(psi1, x.x_is[i].phi1phi2_neg);
}
free_svector(psi1);
psi1 = temp_psi;
for (j=(i+1); j < (x.n_pos+x.n_neg); j++){
if(x.neighbors[i][j]){
if(y.labels[i] != y.labels[j]){
temp_sub = sub_ss_sq(x.x_is[i].phi1phi2_pos, x.x_is[j].phi1phi2_pos);
temp_psi = add_ss(psi2, temp_sub);
free_svector(temp_sub);
free_svector(psi2);
psi2 = temp_psi;
}
}
}
}
// scale w1 by 1/n
temp_psi = smult_s(psi1, (float)1/(float)(x.n_pos+x.n_neg));
free_svector(psi1);
psi1 = temp_psi;
// scale w2 by 1/n^2
if (x.n_neighbors){
temp_psi = smult_s(psi2, (float)1/(float)x.n_neighbors);
free_svector(psi2);
psi2 = temp_psi;
}
// concatenate psi1, psi2
temp_psi = create_svector_with_index(psi2->words, "", 1, (sparm->phi1_size+sparm->phi2_size)*2);
free_svector(psi2);
fvec = add_ss(psi1, temp_psi);
free_svector(temp_psi);
free_svector(psi1);
return(fvec);
}
