ddrous / ncflow

Neural context flow for learning generalisable dynamical systems in multiple environments

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

NCFlow (previously called Nodax and TAX)

Neural Context Flow: A library for Generalisable Neural ODEs

This is a one-stop shop for all things neural odes for dynamical systems (hopefully, WIP).

What to do: Install the library and run the scripts in the folder examples/.

pip install -e .

NCflow is built around 5 extensible modules:

  • a DataLoader: to store the dataset
  • a Learner: a model and loss function
  • A Trainer: to train
  • a VisualTester: to test and visualize the results

A few neural ODE implemented models:

  • One-Per-Env
  • One-For-All
  • Context-Informed
  • Neural Context Flow

About

Neural context flow for learning generalisable dynamical systems in multiple environments

License:MIT License


Languages

Language:Python 96.7%Language:Shell 3.3%