- The paper
- More details in my blog article
visualzing 20newsgroup data
- you can also play with the embedding interactively
label embedding from stackexchange.datascience (deep-learning
as an example)
- you can also play with the embedding interactively
core.py
20newsgroup visualization
20newsgroup_train.py
: train for 20newsgroup dataset20newsgroup_viz.py
: visualization usingsklearn.manifold.TSNE
20newsgroup_tensorboard_embedding.py
: produce the embedding files for tensorboard projector, which is more interactive- you can also play with it here using trained embeddings
link prediction
link_prediction.py
: train (including grid search) and test
stackexchange label visualization
stackexchange_train.py
: train for the stackexchange label cooccurence graphstackexchange_label_embedding.py
: produce the embedding files for tensorboard projector- you can also play with it here using trained embeddings