search

neural-structured-additive-learning / safareg

GNU General Public License v3.0
0 stars 0 forks source link

Change onehot to embedding #2

Open davidruegamer opened 1 year ago

davidruegamer commented 1 year ago

I think we can further improve runtime and memory by using layer_embedding instead of the one-hot encoding. Thx @mb706! onehot_vs_embedding.pdf