A computationally efficient approach to state estimation and learning in graph-structured state-space models with (partially) unknown dynamics and limited historical data.
For more information, see our paper Deep Gaussian Markov Random Fields for Graph-Structured Dynamical Systems
To install the stdgmrf
python package, run
python -m pip install --upgrade build
python -m build
pip install .
To train an ST-DGMRF and perform inference, run
python scripts/run_stdgmrf.py dataset=<dataset-name>
This will use the default settings defined in scripts/conf/config.yaml
. To change these settings, you can either adjust this file, or change them via the command line (will be parsed with hydra).
We use Weights & Biases sweeps to perform our experiments. All relevant config files defining these sweeps can be found in scripts/experiments. To run an experiment, first initialize the sweep with
wandb sweep --project <propject-name> <path-to-config file>
and then, using the obtained agent-ID
, run
wandb agent <agent-ID>