Bruce Ho

factor initialization

2

Any one see the logic in how the factors were initialized? apparently using QR decomposition to obtain 2 orthogonal matrices, dot product them by the lower of the two matrix...

this code uses old version chainer (1.5). its so old, I can't find an instance of ubuntu that can install the corresponding version of cuda. Has anyone attempted to update...

is the recommended way of saving and loading a trained model simply using pickle? or is there a fault proof way of saving only the components like the weights, topics,...

sample code for finding topic distribution for a test dataset

1

Hi, can you provide some sample code for applying this topic model to an arbitrary test dataset to get its topic distribution?

is the 'additional_answers' field from the original training data ever used for anything?

bug in extract_POS()

1

I finally got the code working. The extract_POS code in main_NER.py has this logic which picks up a start_pos. It ends up making terms_arr blank after that point, and nothing...

how to reproduce the pre-trained scoring functions

1

Hi super work! Congratulations for establishing a new state of the art. You checked in pre-trained scoring functions like data/jnk3/jnk3.pkl. It looks like they are RandomForestClassifiers, even though figure 2...

If I understand correctly, the encoder is there to implement the Variational Inference term for enhancing entropy of the feature space. The latent z variable is represented by diagonal Gaussian....

Hi what's the updated url for http://kimi.ml.cmu.edu/mnist.True.3.best.pixel. This one is dead.

I cloned the repo and made the modifications to run with python 3.6. When running mnist_train.py, I get this error RuntimeError: The expanded size of the tensor (3) must match...