tfkld

Code for the EMNLP 2013 paper

• computeWeight.m - read the sparse matrix of the training set and compute the KLD weights
• nmfLS2 - Non-negative matrix factorization with sparse matrix
• nndSVD - NNDSVD algorithm for initializing W and H in NMF

@inproceedings{Ji:2013:EMNLP,
    title = {Discriminative Improvements to Distributional Sentence Similarity},
    author = {Yangfeng Ji and Jacob Eisenstein}
    booktitle = {EMNLP},
    year = {2013},
}