non-negative PU Learning (nnPULearning)

unofficial pytorch reimplementation of nnPULearning. Official chainer implementation is here.

$ python train.py
$ tensorboard --logdir ./result

TODO:

• fix evaluation (calculates mean score by considering batch size)