Closed-form Continuous-time Neural Networks (CfCs) are powerful sequential neural information processing units.
Paper Open Access: https://www.nature.com/articles/s42256-022-00556-7
Arxiv: https://arxiv.org/abs/2106.13898
pip install cfc-model
- Python 3.7 or newer
- Tensorflow 2.4 or newer
- Pandas
- Numpy
For a fresh anaconda environment with the required dependencies:
conda env create --file environment.yml
conda activate cfc
from cfc_model.dense_models import SequentialModel
import numpy as np
X = np.array([[1, 1, 1, 0], [1, 1, 0, 1], [1, 0, 0, 1], [1, 1, 0, 0],
[1, 0, 1, 0], [1, 1, 0, 1], [1, 0, 0, 1], [1, 0, 1, 0]])
y = np.array([0, 0, 1, 1, 1, 0, 1, 1])
model = SequentialModel()
model.fit(X, y)
y_pred = model.predict([1, 1, 0, 1]) # y_pred equals 0
The following configuration states can be used
no_gate
Runs the CfC without the (1-sigmoid) partminimal
Runs the CfC direct solutionuse_ltc
Runs an LTC with a semi-implicit ODE solver instead of a CfCuse_mixed
Mixes the CfC's RNN-state with a LSTM to avoid vanishing gradients
If none of these flags are provided, the full CfC model is used
# Runs an LTC with a semi-implicit ODE solver instead of a CfC
config = {"use_ltc": True}
model.fit(X, y, config=config)
@article{hasani_closed-form_2022,
title = {Closed-form continuous-time neural networks},
journal = {Nature Machine Intelligence},
author = {Hasani, Ramin and Lechner, Mathias and Amini, Alexander and Liebenwein, Lucas and Ray, Aaron and Tschaikowski, Max and Teschl, Gerald and Rus, Daniela},
issn = {2522-5839},
month = nov,
year = {2022}