OLMM

Reference implementation of the Orthogonal Linear Mixing Model (OLMM)

Installation

Python 3.6 or higher is required. To begin with, clone and enter the repo.

git clone https://github.com/wesselb/olmm
cd olmm

Then make a virtual environment and install the requirements.

virtualenv -p python3 venv
source venv/bin/activate
pip install -r requirements.txt

Finally, download the data for the example, which consists of daily temperature measurements across Croatia from 2006.

sh fetch_data.sh

Reference Implementation

A basic reference implementation of the OLMM can be found in olmm.py. It illustrates how to do training, inference, and prediction. These functions are used with AutoGrad, TensorFlow, and PyTorch as backend in example_autograd.py, example_tensorflow.py, and example_pytorch.py, respectively. The three examples execute the same task of loading and preparing the data, fitting a simple OLMM, making predictions and plotting some of the latent processes and some of the outputs.