https://github.com/carpedm20/NTM-tensorflow

"Neural Turing Machine" in Tensorflow
https://github.com/carpedm20/NTM-tensorflow

neural-turing-machines tensorflow

Last synced: 11 days ago
JSON representation

"Neural Turing Machine" in Tensorflow

• Host: GitHub
• URL: https://github.com/carpedm20/NTM-tensorflow
• Owner: carpedm20
• License: mit
• Created: 2015-12-09T14:22:21.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2017-05-18T06:31:02.000Z (over 7 years ago)
• Last Synced: 2024-10-15T22:41:20.910Z (24 days ago)
• Topics: neural-turing-machines, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 33.6 MB
• Stars: 1,046
• Watchers: 67
• Forks: 214
• Open Issues: 11
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-tensorflow - Neural Turing Machine in TensorFlow - implementation of Neural Turing Machine (Models/Projects)
• Awesome-TensorFlow-Chinese - Neural Turing Machine in TensorFlow - implementation of Neural Turing Machine (模型项目 / 微信群)
• awesome-tensorflow - Neural Turing Machine in TensorFlow - implementation of Neural Turing Machine (Models/Projects)
• fucking-awesome-tensorflow - Neural Turing Machine in TensorFlow - implementation of Neural Turing Machine (Models/Projects)

README

        
Neural Turing Machine in Tensorflow

===================================

Tensorflow implementation of [Neural Turing Machine](http://arxiv.org/abs/1410.5401). This implementation uses an LSTM controller. NTM models with multiple read/write heads are supported.

![alt_tag](etc/NTM.gif)

The referenced torch code can be found [here](https://github.com/kaishengtai/torch-ntm).

** 1. Loss sometimes goes to nan even with the gradient clipping ([#2](https://github.com/carpedm20/NTM-tensorflow/issues/2)).**

** 2. The code is very poorly design to support NTM inputs with variable lengths. Just use this code as a reference.**

Prerequisites

-------------

- Python 2.7 or Python 3.3+

- [Tensorflow 1.1.0](https://www.tensorflow.org/)

- NumPy

Usage

-----

To train a copy task:

    $ python main.py --task copy --is_train True

To test a *quick* copy task:

    $ python main.py --task copy --test_max_length 10

Results

-------

More detailed results can be found [here](ipynb/NTM\ Test.ipynb).

**Copy task:**

![alt_tag](etc/result4.png)

![alt_tag](etc/result3.png)

**Recall task:**

(in progress)

Author

------

Taehoon Kim / [@carpedm20](http://carpedm20.github.io/)