This is the companion code for the dynamics model learning method reported in the paper Probabilistic Recurrent State-Space Models by Andreas Doerr et al., ICML 2018. The paper can be found here https://arxiv.org/abs/1801.10395. The code allows the users to reproduce the PR-SSM results reported in the benchmark and large-scale experiments. Please cite the above paper when reporting, reproducing or extending the results.
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
The PR-SSM code depends on Tensorflow.
In order to train a PR-SSM model for a new dataset, a new task has to be derived from the task base class. See for example real_world_tasks.py.
A valid path must be provided to store the experimental results and log files. An example is given in run_benchmark_experiments.py.
The experiments reported in the publication can be run by executing
python benchmarks/run_real_world_tasks/run_benchmark_experiments.py
python benchmarks/run_real_world_tasks/run_large_scale_experiment.py
The individual datasets have to be provided in the datasets folder.
Probabilistic recurrent state-space models is open-sourced under the MIT license. See the LICENSE file for details.
For a list of other open source components included in Benchmarks, see the file 3rd-party-licenses.txt.