Cora dataset contains a citation gragh which also contains feature vectors for each paper. The size of the feature vectors is 1433 (corresponding to unique words). Feature vectors contain binary values indicating whether each word in the vocabulary is present (indicated by 1) or absent (indicated by 0) in the paper.

When I look at these feature vectors only 1 among 1433 positions contain a 1, the rest are 0. I know that node classification in this dataset is done essentially by the citation graph. However, such extremely sparse feature vectors seem rather unusual.

Do you have any explanation for this situation?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.