Deep Mahalanobis Gaussian process

This repository contains all the code needed to replicate the synthetic experiments presented in the Master's degree dissertation "Contributions on latent projections for Gaussian process modeling", avaible at Federal University of Ceará's repository. The code for the model is not neatly packaged into a Python package yet but can be readly imported and used. See any of the Jupyter notebooks for example usage.

Dependencies

See requirements.txt.

How to replicate the experiments

Simply run the main.ipynb Jupyter notebook.

retrain_models variable

The retrain_models variable controls the following behavior:

retrain_models value	Behaviour
True	Run experiments and retrain the models from scratch
False	Load the hyperparameters used in disseration and run the experiments
"save"	Run experiments, retrain the models from scratch, and save the hyperparameters