StochasticForceInference is a package aimed at reconstructing spatially variable force, diffusion and current fields and computing entropy production in Brownian processes. It implements the method presented in SFI_TheoryManual.pdf. Reference:
Learning Force Fields from Stochastic Trajectories, Anna Frishman and Pierre Ronceray, Phys. Rev. X 10, 021009, 2020.
Package author: Pierre Ronceray. Contact: pierre.ronceray@outlook.com. Web page: www.pierre-ronceray.net.
This is a force and diffusion inference package for overdamped systems. For underdamped/inertial systems, check StochasticInertialForceInference: https://github.com/ronceray/StochasticInertialForceInference
Developed in Python 3.6. Dependencies:
-
NumPy, SciPy
-
Optional: Matplotlib (2D plotting, recommended); Mayavi2 (3D plotting)
##Contents:
StochasticForceInference.py: a front-end includer of all classes useful to the user.
SFI_data.py: contains a data wrapper class, StochasticTrajectoryData, which formats trajectories for force and diffusion inference. Also contains a number of plotting routines (of the trajectory, vector fields, tensor fields...). See this file for the different ways to initialize it using data.
SFI_inference.py: implements the core force, velocity and diffusion inference class, StochasticForceInference, that reconstructs the these fields, computes error on this reconstruction, and computes entropy production. Takes as input a StochasticTrajectoryData instance, and inference parameters.
SFI_langevin.py: contains the class OverdampedLangevinProcess, which implements a simple Ito integration of Brownian dynamics, useful for testing the method with known models. It takes as input a force field and a diffusion tensor field. Also used by StochasticForceInference to bootstrap new trajectories with the inferred force field.
SFI_projectors.py: implements an internal class, TrajectoryProjectors, used by DiffusionInference and StochasticForceInference. Given a set of fitting functions, it orthonormalizes it as a premise to the inference.
SFI_bases.py: provides an internal dictionary of (more or less) standard fitting bases, such as polynomials. This dictionary is called by DiffusionInference and StochasticForceInference at initialization, unless a custom base is provided by the user.
SFI_plotting_toolkit.py: a few plotting functions for the convenience of the author.
SFI_demo_Lorenz.py: a fully commented example of force and diffusion inference on the example of a simple 3D process. Start here!
SFI_demo_particles.py: a more advanced example: active Brownian particles. Shows how to extract the pair interaction force.
Enjoy, and please send feedback at pierre.ronceray@outlook.com !
PR