Metadata-Version: 2.1
Name: tmflow
Version: 0.1.0
Summary: Taylor map flow is a package for a 'flowly' construction and learning of polynomial neural networks (PNN) for time-evolving process prediction
Home-page: https://github.com/PNN-Lab/tmflow
Author: PNN Lab
Author-email: golovkina.a@gmail.com
Project-URL: Bug Reports, https://github.com/PNN-Lab/tmflow/issues
Project-URL: Source, https://github.com/PNN-Lab/tmflow
Keywords: PNN,Taylor,ODE,TensorFlow
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Science/Research
Classifier: Intended Audience :: Developers
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Scientific/Engineering :: Mathematics
Classifier: Topic :: Scientific/Engineering :: Physics
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3 :: Only
Requires-Python: >=3.7, <4
Description-Content-Type: text/markdown
License-File: LICENSE

Taylor map flow is a package for a 'flowly' construction and learning of polynomial neural networks (PNN) for time-evolving process prediction.

Based on the input time-series data, it provides:
  - (construct) a module to construct ordinary differential equations (ODEs) in the polynomial form
  - (map) a module to construct a matrix Taylor map for ODEs
  - (learn) a TensorFlow-based module to build and train a polynomial neural network (PNN).
Taylor map matrices can be used as PNN initial weights.

PNN built in this flow way is strongly connected with ordinary differential equations.
This combination reveals the data-underlying deterministic process without manual equation derivation 
and allows treating cases even when only small datasets or partial measurements are available. 
The proposed hybrid models provide explainable and interpretable results to leverage optimal control applications.

'Construct', 'map', and 'learn' modules can be used sequentially or independently from each other.
