Log-likelihood of Dirichlet process

[vchahun]

I think you got it wrong:

for (auto& dish_loc : dish_locs_)
  lp += lgamma(dish_loc.second.num_tables());

should be:

for (auto& dish_loc : dish_locs_)
  for (auto& bin : dish_loc.second.h[0])
    lp += lgamma(bin.first) * bin.second;

According to my derivation:

(where c_t is the count of table t and l_t its label)

Let me know if I am wrong!

[redpony]

Yes, you are certainly right. Not sure where this came from. :)

On Sat, Apr 27, 2013 at 1:59 AM, Victor Chahuneau [email protected]:

I think you got ithttps://github.com/redpony/cpyp/blob/master/cpyp/crp.h#L296wrong:

for (auto& dish_loc : dish_locs_) lp += lgamma(dish_loc.second.num_tables());

should be:

for (auto& dish_loc : dish_locs_) for (auto& bin : dish_loc.second.h[0]) lp += lgamma(bin.first) * bin.second;

According to my derivation:

[image: image]https://f.cloud.github.com/assets/470022/433900/8fd941b8-aeff-11e2-9531-b288650e6e5b.png

(where c_t is the count of table t and l_t its label)

Let me know if I am wrong!

— Reply to this email directly or view it on GitHubhttps://github.com/redpony/cpyp/issues/1 .