/* compute the mean, the covariance matrix, and the eigenvectors.
They are stored in the structure itself */
void pca_online_complete (struct pca_online_s * pca)
{
int d = pca->d;
int n = pca->n;
fvec_div_by (pca->mu, d, n);
fvec_div_by (pca->cov, d * d, n);
float * mumut = fvec_new (d*d);
fmat_mul_tr (pca->mu, pca->mu, d, d, 1, mumut);
fvec_sub (pca->cov, mumut, d*d);
free (mumut);
assert(fvec_all_finite(pca->cov,d*d));
pca->eigvec = fmat_new_pca_from_covariance (d, pca->cov, pca->eigval);
}
float *fmat_center_columns(int d,int n,float *v)
{
assert(n>0);
float *accu=fvec_new_cpy(v,d);
long i;
for(i=1;i<n;i++)
fvec_add(accu,v+i*d,d);
fvec_div_by(accu,d,n);
for(i=0;i<n;i++)
fvec_sub(v+i*d,accu,d);
return accu;
}
/* compute the mean and covariance matrix */
void pca_online_cov (struct pca_online_s * pca)
{
int d = pca->d;
int n = pca->n;
fvec_div_by (pca->mu, d, n);
fvec_div_by (pca->cov, d * (long)d, n);
float * mumut = fvec_new (d*(long)d);
fmat_mul_tr (pca->mu, pca->mu, d, d, 1, mumut);
fvec_sub (pca->cov, mumut, d*(long)d);
free (mumut);
fvec_mul_by (pca->cov, d * (long)d, n / (double) (n-1));
assert(fvec_all_finite(pca->cov,d*(long)d));
pca->n = -pca->n;
}
/*
* A stres test for the addition of feature vectors
*/
int test_stress_add()
{
int i, j, err = 0;
fvec_t *fx, *fy, *fz;
char buf[STR_LENGTH + 1];
test_printf("Stress test for addition of feature vectors");
/* Create empty vector */
fz = fvec_extract("aa0bb0cc", 8, "zero");
for (i = 0; i < NUM_VECTORS; i++) {
/* Create random key and string */
for (j = 0; j < STR_LENGTH; j++)
buf[j] = rand() % 10 + '0';
buf[j] = 0;
/* Extract features */
fx = fvec_extract(buf, strlen(buf), "test");
/* Add fx to fz */
fy = fvec_add(fz, fx);
fvec_destroy(fz);
err += fabs(fvec_norm2(fy) - 1.4142135623) > 1e-7;
/* Substract fx from fz */
fz = fvec_sub(fy, fx);
fvec_sparsify(fz);
/* Clean up */
fvec_destroy(fy);
fvec_destroy(fx);
}
fvec_destroy(fz);
test_return(err, i);
return err;
}
void fmat_subtract_from_columns(int d,int n,float *v,const float *avg) {
long i;
for(i=0;i<n;i++)
fvec_sub(v+i*d,avg,d);
}
/* estimate the GMM parameters */
static void gmm_compute_params (int n, const float * v, const float * p,
gmm_t * g,
int flags,
int n_thread)
{
long i, j;
long d=g->d, k=g->k;
float * vtmp = fvec_new (d);
float * mu_old = fvec_new_cpy (g->mu, k * d);
float * w_old = fvec_new_cpy (g->w, k);
fvec_0 (g->w, k);
fvec_0 (g->mu, k * d);
fvec_0 (g->sigma, k * d);
if(0) {
/* slow and simple */
for (j = 0 ; j < k ; j++) {
double dtmp = 0;
for (i = 0 ; i < n ; i++) {
/* contribution to the gaussian weight */
dtmp += p[i * k + j];
/* contribution to mu */
fvec_cpy (vtmp, v + i * d, d);
fvec_mul_by (vtmp, d, p[i * k + j]);
fvec_add (g->mu + j * d, vtmp, d);
/* contribution to the variance */
fvec_cpy (vtmp, v + i * d, d);
fvec_sub (vtmp, mu_old + j * d, d);
fvec_sqr (vtmp, d);
fvec_mul_by (vtmp, d, p[i * k + j]);
fvec_add (g->sigma + j * d, vtmp, d);
}
g->w[j] = dtmp;
}
} else {
/* fast and complicated */
if(n_thread<=1)
compute_sum_dcov(n,k,d,v,mu_old,p,g->mu,g->sigma,g->w);
else
compute_sum_dcov_thread(n,k,d,v,mu_old,p,g->mu,g->sigma,g->w,n_thread);
}
if(flags & GMM_FLAGS_1SIGMA) {
for (j = 0 ; j < k ; j++) {
float *sigma_j=g->sigma+j*d;
double var=fvec_sum(sigma_j,d)/d;
fvec_set(sigma_j,d,var);
}
}
long nz=0;
for(i=0; i<k*d; i++)
if(g->sigma[i]<min_sigma) {
g->sigma[i]=min_sigma;
nz++;
}
if(nz) printf("WARN %ld sigma diagonals are too small (set to %g)\n",nz,min_sigma);
for (j = 0 ; j < k ; j++) {
fvec_div_by (g->mu + j * d, d, g->w[j]);
fvec_div_by (g->sigma + j * d, d, g->w[j]);
}
assert(finite(fvec_sum(g->mu, k*d)));
fvec_normalize (g->w, k, 1);
printf ("w = ");
fvec_print (g->w, k);
double imfac = k * fvec_sum_sqr (g->w, k);
printf (" imfac = %.3f\n", imfac);
free (vtmp);
free (w_old);
free (mu_old);
}
void vlad_compute(int k, int d, const float *centroids, int n, const float *v,int flags, float *desc)
{
int i,j,l,n_quantile,i0,i1,ai,a,ma,ni;
int *perm ;
float un , diff;
float *tab,*u,*avg,*sum,*mom2,*dists;
int *hist,*assign;
if(flags<11 || flags>=13)
{
assign=ivec_new(n);
nn(n,k,d,centroids,v,assign,NULL,NULL);
if(flags==6 || flags==7)
{
n_quantile = flags==6 ? 3 : 1;
fvec_0(desc,k*d*n_quantile);
perm = ivec_new(n);
tab = fvec_new(n);
ivec_sort_index(assign,n,perm);
i0=0;
for(i=0;i<k;i++)
{
i1=i0;
while(i1<n && assign[perm[i1]]==i)
{
i1++;
}
if(i1==i0) continue;
for(j=0;j<d;j++)
{
for(l=i0;l<i1;l++)
{
tab[l-i0]=v[perm[l]*d+j];
}
ni=i1-i0;
fvec_sort(tab,ni);
for(l=0;l<n_quantile;l++)
{
desc[(i*d+j)*n_quantile+l]=(tab[(l*ni+ni/2)/n_quantile]-centroids[i*d+j])*ni;
}
}
i0=i1;
}
free(perm);
free(tab);
}
else if(flags==5)
{
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
desc[assign[i]*d+j]+=v[i*d+j];
}
}
}
else if(flags==8 || flags==9)
{
fvec_0(desc,k*d);
u = fvec_new(d);
for(i=0;i<n;i++)
{
fvec_cpy(u,v+i*d,d);
fvec_sub(u,centroids+assign[i]*d,d);
un=(float)sqrt(fvec_norm2sqr(u,d));
if(un==0) continue;
if(flags==8)
{
fvec_div_by(u,d,un);
} else if(flags==9)
{
fvec_div_by(u,d,sqrt(un));
}
fvec_add(desc+assign[i]*d,u,d);
}
free(u);
}
else if(flags==10)
{
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
desc[assign[i]*d+j]+=v[i*d+j];
}
}
for(i=0;i<k;i++)
{
fvec_normalize(desc+i*d,d,2.0);
}
}
else if(flags==13)
{
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
desc[assign[i]*d+j]+=(float)sqr(v[i*d+j]-centroids[assign[i]*d+j]);
}
}
}
else if(flags==14)
{
avg = fvec_new_0(k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
avg[assign[i]*d+j]+=v[i*d+j]-centroids[assign[i]*d+j];
}
}
hist=ivec_new_histogram(k,assign,n);
for(i=0;i<k;i++)
{
if(hist[i]>0)
{
for(j=0;j<d;j++)
{
avg[i*d+j]/=hist[i];
}
}
}
free(hist);
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
desc[assign[i]*d+j]+=(float)(sqr(v[i*d+j]-centroids[assign[i]*d+j]-avg[assign[i]*d+j]));
}
}
fvec_sqrt(desc,k*d);
free(avg);
}
else if(flags==15)
{
fvec_0(desc,k*d*2);
sum = desc;
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
sum[assign[i]*d+j]+=v[i*d+j]-centroids[assign[i]*d+j];
}
}
hist = ivec_new_histogram(k,assign,n);
mom2 = desc+k*d;
for(i=0;i<n;i++)
{
ai=assign[i];
for(j=0;j<d;j++)
{
mom2[ai*d+j]+=(float)(sqr(v[i*d+j]-centroids[ai*d+j]-sum[ai*d+j]/hist[ai]));
}
}
fvec_sqrt(mom2,k*d);
free(hist);
}
else if(flags==17)
{
fvec_0(desc,k*d*2);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
diff=v[i*d+j]-centroids[assign[i]*d+j];
if(diff>0)
{
desc[assign[i]*d+j]+=diff;
}
else
{
desc[assign[i]*d+j+k*d]-=diff;
}
}
}
}
else
{
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
desc[assign[i]*d+j]+=v[i*d+j]-centroids[assign[i]*d+j];
}
}
if(flags==1)
{
hist=ivec_new_histogram(k,assign,n);
/* printf("unbalance factor=%g\n",ivec_unbalanced_factor(hist,k)); */
for(i=0;i<k;i++)
{
for(j=0;j<d;j++)
{
desc[i*d+j]/=hist[i];
}
}
free(hist);
}
if(flags==2)
{
for(i=0;i<k;i++)
{
fvec_normalize(desc+i*d,d,2.0);
}
}
if(flags==3 || flags==4)
{
assert(!"not implemented");
}
if(flags==16)
{
hist=ivec_new_histogram(k,assign,n);
for(i=0;i<k;i++)
{
if(hist[i]>0)
{
fvec_norm(desc+i*d,d,2);
fvec_mul_by(desc+i*d,d,sqrt(hist[i]));
}
}
free(hist);
}
}
free(assign);
}
else if(flags==11 || flags==12)
{
ma=flags==11 ? 4 : 2;
assign=ivec_new(n*ma);
dists=knn(n,k,d,ma,centroids,v,assign,NULL,NULL);
fvec_0(desc,k*d);
for(i=0;i<n;i++)
{
for(j=0;j<d;j++)
{
for(a=0;a<ma;a++)
{
desc[assign[ma*i+a]*d+j]+=v[i*d+j]-centroids[assign[ma*i+a]*d+j];
}
}
}
free(dists);
free(assign);
}
}
