Hybrid Modeling is a modeling technique in which a physics-based model is combined with a data-based approach. By combining the best of both worlds, we trade off the benefit of prior knowledge when training data is scarce with the flexibility of a data-driven approaches when training data is abundant. In this notebook we demonstrate the power of hybrid modeling by modeling data, which roughly follows the amplitude of a non-linearly damped oscillator with missing data.
The paper Hybrid Modeling Design Patterns (Maja Rudolph, Stefan Kurz, Barbara Rakitsch) provides additional context and background to this notebook, as well as a broader perspective on the topic.
This software is a tutorial, solely developed for educational purpose. It will neither be maintained nor monitored in any way.
First create a conda environment
conda create -n HYM_notebook python=3.11
Activate the environment
conda activate HYM_notebook
Install packages from requirements file using conda
pip install -r requirements.txt
If you have problems, importing the packages in Jupyter, add your Conda environment to Jupyter as a new kernel.
python -m ipykernel install --user --name=HYM_notebook --display-name "HYM notebook"
Open Jupyter, go to Kernel > Change Kernel and select "HYM notebook".
Benchmarks is open-sourced under the MIT license. See the LICENSE file for details.