PyRates -- A Code-Generation Tool for Dynamical Systems Modeling

TL;DR

PyRates is a Python-based tool that simplifies the modeling and analysis of dynamical systems by providing a user-friendly interface and versatile code-generation capabilities across multiple programming languages.

Contribution

It introduces a hierarchical, graph-based modeling framework and a code-generation system that translates models into various languages, enhancing flexibility and reproducibility in dynamical systems research.

Findings

Successfully generates code in Python, Fortran, and Julia.

Demonstrates applications in numerical simulation, bifurcation analysis, and neural network optimization.

Supports development of new dynamical systems tools.

Abstract

Mathematical models allow us to gain a deeper understanding of real-world dynamical systems. One of the most powerful mathematical frameworks for modeling real-world phenomena are systems of differential equations. In the majority of fields that use differential equations, numerical methods are essential for conducting model-based research. Although many software solutions are available for the numerical study of differential equation systems, a common framework for implementing differential equation systems is lacking. This hinders progress in dynamical systems research and limits the shareability and reproducibility of results. PyRates is a Python-based software for modeling and analyzing dynamical systems. It provides a user-friendly interface for defining models, which is based on a graph-based, hierarchical structure that mirrors the modular organization of real-world dynamical systems. This design allows users to leverage the hierarchical structure of their systems and create their models with minimal effort. Importantly, the core of PyRates is a versatile code-generation system, which can translate user-defined models into "backend" implementations in various languages, including Python, Fortran, and Julia. This allows users to access a wide range of analysis methods for dynamical systems, eliminating the need for manual translation between code bases. We demonstrate PyRates's capabilities in three use cases, where it generates NumPy code for numerical simulations, Fortran code for bifurcation analysis, and PyTorch code for neural network optimization. Finally, PyRates can be used as a model definition interface for the creation of new dynamical systems tools. We developed two such software packages, PyCoBi and RectiPy, as extensions of PyRates for specific dynamical systems modeling applications.