meereeum / lda2vec-tf Goto Github PK

Hi, Figure 1 in the paper shows that there first comes document weights, then after a softmax layer, there comes document proportions. But in the loss function, there is no document weights, only the document proportion p_jk. So are document proportions parameters in the network, or are they just learned by other methods beforehand? Thanks!

can you give some details what are the expected results and which files we are supposed to run

The loss in your implementation is not exactly the same as in the original paper by Moody.
Am I right or wrong?

Hi, when I have done training a lda2vec model, how can I use this model to infer the vector of a new document. I couldn't find any inference function in lda2vec-tf/model.py.

To get the import working, change line
from tensorflow.contrib.tensorboard.plugins import projector
to
from tensorboard.plugins import projector
in files model.py and save_embeds.py.

Giving same representaion for every document. Has anyone faced the same problem?

Hi,

I tried to visualize the output of your work, sadly it does not work as it needs the embedding file for the visualization,
In none of the steps this file is being created,

exactly this line in save_embeds.py : feed_dict = {embed_vals: np.load(f_embed)}

I dont have the f_embed to pass,

May I have your view on this?
thanks

It will make it easier to understand and use.

After we run the code and get the document embedding how can we use it to predit the embedding of a new unobserved document?

All the topic vectors are same. Is it working for you?

when trying to run preprocess.py I got this error

how to train this model with my specific data stored in mysql database ??