search

commonsense / conceptnet-numberbatch

Other
1.28k stars 143 forks source link

Embedding for other dimensions: 50, 100 and 200 #58

Closed zhongpeixiang closed 5 years ago

zhongpeixiang commented 5 years ago

It's really nice to see concept-enriched embeddings!

It would be nicer to have embeddings in other dimensions, e.g., 50, 100 and 200 because there are many models that use these smaller dimensions to prevent overfitting.

lsorber commented 5 years ago

You could use PCA [1] to reduce the dimensionality to however many dimensions you wish. The embeddings start to degrade around the 200 mark though.

[1] https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html

zhongpeixiang commented 5 years ago

Thanks for pointing out using PAC for dimension reduction :).