Code for reproducing experiments on MNIST and CIFAR-10 in paper "Positive-Unlabeled Learning with Non-Negative Risk Estimator".
- Python 3.6.1
- tensorflow 1.4
- matplotlib
- numpy
- sklearn
-
MNIST experiment:
python train.py mnist 1
-
CIFAR-10 experiment:
python train.py cifar10 1
When they are done, a figure will show up. And results are saved in the folder ./result
, then you can see the figures if you run python show_result.py mnist
or python show_result.py cifar10
.
- Result of MNIST experiment:
- Result of CIFAR-10 experiment: