ml_adp is a Python package embedding into the Pytorch neural network and optimization framework and serves the numerical solution of Markovian discrete-time finite-horizon stochastic optimal control problems.
It exports a list-like interface to the central functional components of such optimal control problems, allowing for concise implementations of numerical methods that rely on the approximate satisfaction of the discrete-time Bellman equations (Approximate Dynamic Programming; ADP).
Get Python ~= 3.7 and pip-install the repo to your environment env.
For example:
(env) ➜ pip install git+https://github.com/rwlmu/ml-adp
To use ml_adp in Jupyter notebooks install the IPython kernel dependencies to the environment and create the kernel from within the environment
(env) ➜ pip install "ml_adp[jupyter]"
(env) ➜ python -m ipykernel install --name kernelname
Now, select the kernel kernelname in your Jupyter application instance.
Documentation is available here.