void col_sum(gsl_matrix* m, gsl_vector* val)
{
int i, j;
gsl_vector_set_all(val, 0);
for (i = 0; i < m->size1; i++)
for (j = 0; j < m->size2; j++)
vinc(val, j, mget(m, i, j));
}
/*
* compute the column sums of a matrix
*
*/
void col_sum(gsl_matrix* m, gsl_vector* val) {
gsl_vector_set_all(val, 0);
for (unsigned int i = 0; i < m->size1; i++) {
for (unsigned int j = 0; j < m->size2; j++) {
vinc(val, j, mget(m, i, j));
}
}
}
void update_gamma(lda_post* p) {
int k, n, K = p->model->ntopics, N = p->doc->nterms;
gsl_vector_memcpy(p->gamma, p->model->alpha);
for (n = 0; n < N; n++)
{
gsl_vector phi_row = gsl_matrix_row(p->phi, n).vector;
int count = p->doc->count[n];
for (k = 0; k < K; k++)
vinc(p->gamma, k, vget(&phi_row, k) * count);
}
}
void initialize_lda_ss_from_data(corpus_t* data, lda_suff_stats* ss) {
int k, n, i, w;
gsl_rng * r = new_random_number_generator();
for (k = 0; k < ss->topics_ss->size2; k++)
{
gsl_vector topic = gsl_matrix_column(ss->topics_ss, k).vector;
for (n = 0; n < LDA_SEED_INIT; n++)
{
int d = floor(gsl_rng_uniform(r) * data->ndocs);
doc_t* doc = data->doc[d];
for (i = 0; i < doc->nterms; i++)
{
vinc(&topic, doc->word[n], doc->count[n]);
}
}
for (w = 0; w < topic.size; w++)
{
vinc(&topic, w, LDA_INIT_SMOOTH + gsl_rng_uniform(r));
}
}
}
double c_ctr::doc_inference(const c_document* doc, const gsl_vector* theta_v,
const gsl_matrix* log_beta, gsl_matrix* phi,
gsl_vector* gamma, gsl_matrix* word_ss,
bool update_word_ss) {
double pseudo_count = 1.0;
double likelihood = 0;
gsl_vector* log_theta_v = gsl_vector_alloc(theta_v->size);
gsl_vector_memcpy(log_theta_v, theta_v);
vct_log(log_theta_v);
int n, k, w;
double x;
for (n = 0; n < doc->m_length; n ++) {
w = doc->m_words[n];
for (k = 0; k < m_num_factors; k ++)
mset(phi, n, k, vget(theta_v, k) * mget(m_beta, k, w));
gsl_vector_view row = gsl_matrix_row(phi, n);
vnormalize(&row.vector);
for (k = 0; k < m_num_factors; k ++) {
x = mget(phi, n, k);
if (x > 0)
likelihood += x*(vget(log_theta_v, k) + mget(log_beta, k, w) - log(x));
}
}
if (pseudo_count > 0) {
likelihood += pseudo_count * vsum(log_theta_v);
}
gsl_vector_set_all(gamma, pseudo_count); // smoothing with small pseudo counts
for (n = 0; n < doc->m_length; n ++) {
for (k = 0; k < m_num_factors; k ++) {
x = doc->m_counts[n] * mget(phi, n, k);
vinc(gamma, k, x);
if (update_word_ss) minc(word_ss, k, doc->m_words[n], x);
}
}
gsl_vector_free(log_theta_v);
return likelihood;
}
