This repository contains the python scripts used in our paper published in the Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS) 2021, San Diego, California, USA.
In this work, we propose an unsupervised generative model that can learn a low-dimensional representation among the observations in a latent space, while making use of all available data in a heterogeneous data setting with missing values. We improve upon the existing Gaussian process latent variable model (GPLVM) by incorporating multiple likelihoods and deep neural network parameterised back-constraints to create a non-linear dimensionality reduction technique for heterogeneous data. In addition, we develop a variational inference method for our model that uses numerical quadrature.
Each point in the above image corresponds to an observation comprising of measurements from different likelihoods, that has been projected on to a two dimensional latent space. The colour coding in the latent space would correspond to cluster membership.
These scripts require the following software:
- Python (>= 2.6.0)
- Theano(>= 0.9.0) and associated dependencies
- Create results folder.
- Save the data files in the data folder.
- Update the necessary parameters in the run_model.py file.
- To train the model, run:
python run_model.py
Please cite this work as:
Ramchandran, S., Koskinen, M., & Lähdesmäki, H. (2021). Latent Gaussian process with composite likelihoods and numerical quadrature. Proceedings of the Twenty Fourth International Conference on Artificial Intelligence and Statistics (AISTATS)
This project is licensed under the MIT License - see the LICENSE file for details.