Implementation of Khan and Lin's Conjugate-Computation Variational Inference: Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models for linear dynamical systems w/ poisson likelihood.

Before running our code, create a conda environment using the file environment.yml. To do so, open a terminal and run: conda env create -f environment.yml

Then, activate the created environment: source activate cvi-poisson-lds

If you don't want to use conda, just make sure to use the libraries listed in environment.yml in their specified version. Additionally, the latest versions of Lasagna and Theano are required to generate the data. They are not available on PyPI:

pip install --upgrade https://github.com/Theano/Theano/archive/master.zip
pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip

python main.py

Hyperparameters can be tuned on lines 88 - 96. To use CVI for your own problem, write your own non-conjugate likelihood in cvi_helpers.non_conjugate_likelihood().

Plots are saved in CVI_PLDS/plots (see screenshot below).

We use the synthetic data code and the VILDS model from Black Box Variational Inference for Linear Dynamical Systems by Archer et al.

If you use the code, please cite the original paper. Bibtex:

@article{khan2017conjugate,
  title={Conjugate-Computation Variational Inference: Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models},
  author={Khan, Mohammad Emtiyaz and Lin, Wu},
  journal={arXiv preprint arXiv:1703.04265},
  year={2017}
}