DynX is a computatioanl framework for representing conjugate pairs of functional operations, a push-forward and pull-back, on Banch space as a graph. The focus of is on a practical rather than theoretical representation of Bellman functional recursions and push-forward measures.
The use of graphs and networks to represent dynamics is ubiquitous (see Auclert et al (2021) or Stachurski and Sargent (2025), for instance). However, the critical innovation in DynX is to represent relations between functions and measures, rather than relations between variables. Moreover, Unlike general graph libraries (e.g. TensorFlow), DynX represents both computational operators and the functional objects that define a recursive problem.
DynX is not yet published on PyPI. You can install the latest development build directly from GitHub:
Main branch (bleeding-edge)
pip install "git+https://github.com/akshayshanker/dynx.git#egg=dynx"Specific dev release (v0.18.dev4)
pip install "git+https://github.com/akshayshanker/dynx.git@v0.18.dev4#egg=dynx"To upgrade an existing installation:
pip install --upgrade --force-reinstall \
"git+https://github.com/akshayshanker/dynx.git#egg=dynx"Comprehensive documentation is in progress.