leihuang / infotopo

A Python package for information-geometric and information-topological analysis of mathematical models

infotopo

A Python package for carrying out some analysis of "typical"¹ mathematical models, including:

• parameter estimation
• sampling of parameter posterior distribution
• model reduction
• model selection

The frameworks and methods here follow the treatments of information geometry and its extension information topology², hence the name of the package.

Note 1: The technical definition of "typical" here is any mathematical model whose predictions are differentiable with respect to parameters, which includes most models in physical sciences.

Note 2: See http://doi.org/10.1103/PhysRevE.83.036701 and http://arxiv.org/abs/1409.6203.

Prerequisites

Usage examples

mod = Model()
expts = Experiments()

pred = mod.get_pred(expts)

s = pred.get_spectrum()

dat = pred.get_data()

res = residual.Residual(pred, dat)

fit = fitting.leverberg_marquardt(res, p0)

ens = sampling.sampling(res, p0, nstep=100)

ens.scatterplot()

gds = pred.get_geodesic()

gds.integrate()

gds.plot()