https://github.com/francesco-mannella/echo-state-networks
Echo State Networks with TensorFlow
https://github.com/francesco-mannella/echo-state-networks
esn ipynb python tensorflow
Last synced: 8 months ago
JSON representation
Echo State Networks with TensorFlow
- Host: GitHub
- URL: https://github.com/francesco-mannella/echo-state-networks
- Owner: francesco-mannella
- Created: 2017-10-18T10:09:18.000Z (over 8 years ago)
- Default Branch: master
- Last Pushed: 2020-06-28T05:39:10.000Z (almost 6 years ago)
- Last Synced: 2025-02-01T10:41:32.931Z (over 1 year ago)
- Topics: esn, ipynb, python, tensorflow
- Language: Jupyter Notebook
- Homepage:
- Size: 20.5 MB
- Stars: 37
- Watchers: 8
- Forks: 18
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Implementing Echo-State Networks in tensorflow2.
[ESN-definition.ipynb](ESN-definition.ipynb) Describes the mathematics of leaky echo-state
networks together with an analytical method to correct the spectral radius of the inner weights
accounting for leakiness and a matricial trick to improve the variance of network states
during its dynamics.
[ESN.py](ESN.py) Contains the definition of a customized tensorflow Cell.
The inizialization of the weights uses numpy function via [tf.py_function](https://www.tensorflow.org/api_docs/python/tf/py_function)
because tf.self_adjoint_eigvals only works on self-adjoint matrices.
[ESN-usage.ipynb](ESN-usage.ipynb) Contains an example of training on a simple dataset.
[parametric-sequence-learning.ipynb](parametric-sequence-learning.ipynb) Contains a more complex example. A set of 2D trajectories is learned
and generalization to the whole family of trajectories is tested.