Inference for topic models

Posted: 2016-05-18 , Modified: 2016-05-18

gen_topic_vector: Generate row vectors of size \(K\) using Dirichlet (or Gaussian) \((\al,\ldots, \al)\).
gen_doc: Generate topic vector, multiply by A
interleave
singlen1: \(i\)th standard basis vector
process_row_delta: ?
process_row_C: ?
new_compute_B:
threshold:
log_likelihood:
fast_grad_log_likelihood:
project
l1_distance
projected_ascent: ?
top
top5
is_anchorword
near_anchorword
sparse_topic
superflat_topic
top_likelihood_support
test_inference
safe_mkdir
write_doc
write_flat
estimate_support
estimate_support_map
split_heldout
heldout_score
held_out_probability_ascent
nearest_neighbors
get_train_data

Inference for topic models - Code