https://github.com/mitmul/chainer-mdn

Mixture Density Network with Chainer
https://github.com/mitmul/chainer-mdn

chainer generative-model machine-learning mixture-density-networks numpy

Last synced: 21 days ago
JSON representation

Mixture Density Network with Chainer

• Host: GitHub
• URL: https://github.com/mitmul/chainer-mdn
• Owner: mitmul
• License: mit
• Created: 2018-06-12T15:01:12.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-06-12T18:37:45.000Z (over 6 years ago)
• Last Synced: 2024-11-24T09:41:58.658Z (3 months ago)
• Topics: chainer, generative-model, machine-learning, mixture-density-networks, numpy
• Language: Python
• Size: 462 KB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Mixture Density Networks with Chainer

This is an example code of Mixture Density Networks implemented with Chainer.

All the parameters, the network design, and how to prepare synthetic data are the same as this article: [Mixture Density Networks with TensorFlow](http://blog.otoro.net/2015/11/24/mixture-density-networks-with-tensorflow/).

This repository is intended to be a draft for a repository in [chainer/models](https://github.com/chainer/models) which is to collect self-contained Chainer examples.

## Requirements

- [Chainer](https://chainer.org) v4.1.0

- [NumPy](http://www.numpy.org/) v1.14.0

- [Matplotlib](https://matplotlib.org/) v2.1.2

## Train

```

python train.py

```

## Results

### Target data:

![target data](images/target.png)

### Generated data from the trained MDN:

![generated data](images/generated.png)

### Generated data plotted over the target data:

![overlapped](images/overlap.png)

### Loss curve

![loss](images/loss.png)