deep GAN experiment.
Multiple generators cooperate to improve generative ability under the competitive selection of one discriminator
- tensorflow
- numpy
Step 1. Clone this repository and adjust the environment setting in main.py if necessary.
Ensure your system is installed with Git and clone this reposity with command line:
$ git clone https://github.com/naturomics/deepGAN.git
cd deepGAN and edit file 'main.py' to suit your configuration.
Step 2. Download MNIST dataset, mv and extract them into data/mnist directory.
$ mkdir -p data/mnist
$ cd data/mnist
$ wget -c http://yann.lecun.com/exdb/mnist/{train-images-idx3-ubyte.gz,train-labels-idx1-ubyte.gz,t10k-images-idx3-ubyte.gz,t10k-labels-idx1-ubyte.gz}
$ gunzip *.gz
$ cd - # go back to project root directory
Step 3. Start to run for training with command line:
$ python main.py --dataset mnist --input_height=28 --output_height=28 --is_train
Experiments were carried out with different hyper parameters theta and beta
The legend for various hyper parameters(theta00 meaning theta=0.0 and no using beta, theta04beta08 i.e. theta=0.4 and beta=0.8, etc.):
d1 total loss:
d2 total loss:
d loss with g1 as fake input:
d loss with g2 as fake input:
g1 loss:
g2 loss:
