# NODEFit
(https://pepy.tech/project/nodefit)
Fit time-series data with a Neural Differential Equation!
This repository contains time-series data fit capabilities using both Neural Ordinary Differential Equations and Neural Stochastic Differential Equations
GPU support is packaged as part of [PyTorch](https://pytorch.org/)
# How to install and execute?
Tested on Python 3.9
Just run
```
pip install nodefit
```
The following program illustrates a basic example
```python
import numpy as np
import torch.nn as nn
from nodefit.constants import DEVICE
from nodefit.neural_ode import NeuralODE
from nodefit.neural_sde import NeuralSDE
###
# DEFINE NETWORKS
###
# Neural ODE parameters
ndim, drift_nhidden, diffusion_nhidden = 2, 10, 2
drift_nn = nn.Sequential(
nn.Linear(ndim+1, drift_nhidden),
nn.Sigmoid(),
nn.Linear(drift_nhidden, ndim)
).double().to(DEVICE)
diffusion_nn = nn.Sequential(
nn.Linear(ndim+1, diffusion_nhidden),
nn.Sigmoid(),
nn.Linear(diffusion_nhidden, ndim)
).double().to(DEVICE)
###
# PROVIDE DATA
###
# Training between data for 0 and 5 seconds
t = np.linspace(0, 5, 10)
# Provide data as list of lists with starting condition
data = np.array([[...]])
###
# FIT USING NEURALODE
###
print('Performing fit using Neural ODE...')
neural_ode = NeuralODE(drift_nn, t, data)
neural_ode.train(2000)
# # Extrapolate the training data to 10 seconds
extra_data = neural_ode.extrapolate(10)
neural_ode.plot(extra_data)
###
# FIT USING NEURALSDE
###
print('Performing fit using Neural SDE...')
neural_sde = NeuralSDE(drift_nn, diffusion_nn, t, data)
neural_sde.train(1)
# # Extrapolate the training data to 10 seconds
extra_data = neural_sde.extrapolate(10)
neural_sde.plot(extra_data)
```
## Sample Output
## Whom to contact?
Please direct your queries to [gpavanb1](http://github.com/gpavanb1)
for any questions.
## Acknowledgements
This package would not be possible without the supporting packages - [torchdiffeq](https://github.com/rtqichen/torchdiffeq) and [torchsde](https://github.com/google-research/torchsde)
Raw data
{
"_id": null,
"home_page": "https://github.com/gpavanb1/NODEFit",
"name": "NODEFit",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "python neural-network pytorch numerical-methods neural-ode",
"author": "gpavanb1",
"author_email": "gpavanb@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/24/7d/1cf1a2cfbcd48378ea6505c38f48a6d5adbab04e29ceb86fd85dc24d3ecc/NODEFit-0.1.tar.gz",
"platform": null,
"description": "# NODEFit\n\n(https://pepy.tech/project/nodefit)\n\nFit time-series data with a Neural Differential Equation!\n\nThis repository contains time-series data fit capabilities using both Neural Ordinary Differential Equations and Neural Stochastic Differential Equations\n\nGPU support is packaged as part of [PyTorch](https://pytorch.org/)\n\n# How to install and execute?\n\nTested on Python 3.9\n\nJust run \n```\npip install nodefit\n```\n\nThe following program illustrates a basic example\n```python\nimport numpy as np\nimport torch.nn as nn\nfrom nodefit.constants import DEVICE\n\nfrom nodefit.neural_ode import NeuralODE\nfrom nodefit.neural_sde import NeuralSDE\n\n\n###\n# DEFINE NETWORKS\n###\n\n# Neural ODE parameters\nndim, drift_nhidden, diffusion_nhidden = 2, 10, 2\n\ndrift_nn = nn.Sequential(\n nn.Linear(ndim+1, drift_nhidden),\n nn.Sigmoid(),\n nn.Linear(drift_nhidden, ndim)\n).double().to(DEVICE)\n\ndiffusion_nn = nn.Sequential(\n nn.Linear(ndim+1, diffusion_nhidden),\n nn.Sigmoid(),\n nn.Linear(diffusion_nhidden, ndim)\n).double().to(DEVICE)\n\n###\n# PROVIDE DATA\n###\n\n# Training between data for 0 and 5 seconds\nt = np.linspace(0, 5, 10)\n# Provide data as list of lists with starting condition\ndata = np.array([[...]])\n\n###\n# FIT USING NEURALODE\n###\nprint('Performing fit using Neural ODE...')\n\nneural_ode = NeuralODE(drift_nn, t, data)\nneural_ode.train(2000)\n\n# # Extrapolate the training data to 10 seconds\nextra_data = neural_ode.extrapolate(10)\nneural_ode.plot(extra_data)\n\n###\n# FIT USING NEURALSDE\n###\nprint('Performing fit using Neural SDE...')\n\nneural_sde = NeuralSDE(drift_nn, diffusion_nn, t, data)\nneural_sde.train(1)\n\n# # Extrapolate the training data to 10 seconds\nextra_data = neural_sde.extrapolate(10)\nneural_sde.plot(extra_data)\n\n```\n\n## Sample Output\n\n\n\n\n## Whom to contact?\n\nPlease direct your queries to [gpavanb1](http://github.com/gpavanb1)\nfor any questions.\n\n## Acknowledgements\n\nThis package would not be possible without the supporting packages - [torchdiffeq](https://github.com/rtqichen/torchdiffeq) and [torchsde](https://github.com/google-research/torchsde)",
"bugtrack_url": null,
"license": "MIT",
"summary": "Fit time-series data with a Neural Differential Equation",
"version": "0.1",
"project_urls": {
"Bug Reports": "https://github.com/gpavanb1/NODEFit/issues",
"Homepage": "https://github.com/gpavanb1/NODEFit",
"Source": "https://github.com/gpavanb1/NODEFit/"
},
"split_keywords": [
"python",
"neural-network",
"pytorch",
"numerical-methods",
"neural-ode"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "247d1cf1a2cfbcd48378ea6505c38f48a6d5adbab04e29ceb86fd85dc24d3ecc",
"md5": "81333145dadd1ab532e00a80f675e242",
"sha256": "c36790b0d374fd156d178e93b72a6d76f5c8f98f56fa0820a43cd1c09f6446d9"
},
"downloads": -1,
"filename": "NODEFit-0.1.tar.gz",
"has_sig": false,
"md5_digest": "81333145dadd1ab532e00a80f675e242",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 5427,
"upload_time": "2023-12-17T10:11:25",
"upload_time_iso_8601": "2023-12-17T10:11:25.716778Z",
"url": "https://files.pythonhosted.org/packages/24/7d/1cf1a2cfbcd48378ea6505c38f48a6d5adbab04e29ceb86fd85dc24d3ecc/NODEFit-0.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-12-17 10:11:25",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "gpavanb1",
"github_project": "NODEFit",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [],
"lcname": "nodefit"
}
JSON
