Evaluate

It provide a python API to evaluate node embedding on multi-label classification with LIBLINEAR.

How to run

Run with the following procedures:

1. compile the code:

sh make.sh

2a) normalize node embedding:

./norm -input emb.txt -output emb.bin -binary 1

2b) binarize node embedding (without normalization):

./to_binary -input emb.txt -output emb.bin

Format of emb.txt:

10312(number of nodes) 128(embedding dimension)
1(node id) 0.3000 0.4000 ...
2 0.5000 0.6000 ...
...

3. evaluate:

python2 test.py --emb emb.bin --vocab vocab.txt --label label.txt --portion 0.1

Format of vocab.txt:

1(node id)
2
...

Format of label.txt:

1(node id) 1(group id)
1 2
2 1
...