Implementation -- write embedding matrix

[KaiWeiChang]

• Change pmf.cpp to write out the embedding matrix
• Combine the embedding matrix and the vocabulary
• check glove.cpp to see how they combine word matrix and context matrix

[KaiWeiChang]

1. generate coocurance matrix

mkdir example2
cp ???/cooccurance.txt example2/train.ratings
cp ???/cooccurance.txt example2/test.ratings
edit example2/mata
wc -l train.ratings (#lines in train.ratings = nonzero of matrix)
find max row and column values

./converter example2

./omp-pmf-train -s 11 -n 12 -f 1 -k 20 example (make sure it writes embedings)

cut -f 1 -d ' ' vocab.txt > words (give you all the words)
paste words example.model > pmf_embedding

eval ...