binzhou-com / gan_distribution_tutorial

A simple tutorial of training GAN to model distribution

gan distribution demo

A simple demo of training GAN to generate samples for Gamma distribution

Requirements

• Python 3.5
• TensorFlow 1.6
• Numpy
• Matplotlib
• Seaborn

Usage

$ python main.py --help

This will illustrate all the arguments you could play with the code.

A simple demo using gan to generate gamma distributions

optional arguments:
  -h, --help            show this help message and exit
  --gpu GPU             gpu to use: 0, 1, 2, 3, 4. Default: None
  --lr LR               learning rate. Default: 1e-4
  --iterations ITERATIONS
                        num of iterations. Default: 2000
  --alpha ALPHA         Gamma alpha. Default: 2.
  --scale SCALE         Gamma beta. Default: .5
  --sample_size SAMPLE_SIZE
                        sample size. Default: 100
  --model_save_dir MODEL_SAVE_DIR
                        directory to save model. Default: checkpoints
  --res_save_dir RES_SAVE_DIR
                        directory to save results. Default: results
  --log_save_dir LOG_SAVE_DIR
                        directory to save logs. Default: logs

###########################################################################

Train the model

Defaultly, the code will train a generator to fit the Gamma(2, 2), you can play with the code by changing the argument --alpha and --beta to model other Gamma distributions. For example:

$ python main.py --alpha=1. --beta=2.

Results and Logs

The code will save a video which illustrates the training procedure in the directory results/ defaultly.

Author

Haibin Yu/ @HeroKillerEver