A toolbox for exact and approximate inference in continuous-time Bayesian networks. This toolbox contains MATLAB implementations of methods for exact and approximate inference for Continuous-time Bayesian networks (CTBNs) as described in "D. Linzner and H. Koeppl, Cluster Variational Approximations for Structure Learning of Continuous-Time Bayesian Networks from Incomplete Data. Advances in Neural Information Processing Systems 31, 7890--7900,2018".
We provide a couple of test scripts in the folder SCRIPTS for posterior inference and structure learning. With the exception of the IRMA test script (in SynthGRN_IRMA folder) these experiments are performed on synthetic data. For the IRMA test script, data is available at:
http://www.cell.com/supplemental/S0092-8674(09)00156-1
Reference: I. Cantone et al., A Yeast Synthetic Network for In Vivo Assessment of Reverse-Engineering and Modeling Approaches, Cell, vol. 137, no. 1, pp. 172--181, 2009.
We provide a pre-processing script in SynthGRN_IRMA, implementing the observation model in the supplementary material of our manuscript.
Our structure learning method takes in data of the form of two cell-arrays: likelihood of latent state given the observation model: Z=cell{number of trajectory}(state,time-point,node index) Time-points of measurement: TZ=cell{number of trajectory}(time-point)
IMPORTANT: Prior-rates for structure learning have to match time-scales of data provided. Otherwise our method will reject samples with the warning "Warning: Could not process data-point x of node y". We suggest re-scaling data to ~1 Transition per time unit and assuming a rate prior with alpha/beta ~ 1
This toolbox depends in parts on the MATLAB statistics toolbox. For parallel computing MATLABs parallel computing toolbox is needed.
Acknowledgements: Dominik Linzner is funded by the European Union's Horizon 2020 research and innovation programme under grant agreement 668858 (PrECISE).
-Dominik Linzner (dominik.linzner@bcs.tu-darmstadt.de)