内置模块用于Python中的贝叶斯曲线拟合

Question

I was searching scipy library for any built in modules for Bayesian curve fitting and I'm not able to find one. All I found is : scipy.optimize.curve_fit

But the description of this link says that this is non linear least squares fit. My question is - do we have to implement our own module for Bayesian curve fitting or is there any such module that I might have missed?

Answer 1

Bayesian inference is not part of the SciPy library - it is simply out of scope for scipy . There is a number of separate python modules that deal with it, and it seems that you have indeed missed quite a few of those - most notably implementations of Markov chain Monte Carlo algorithms pymc and emcee that are probably the most used MCMC packages. They are both relatively straightforward to set up, but in my opinion emcee is easier to get started with.

As with everything, the devil is in the details with Bayesian curve fitting - I highly recommend reading through this overview to get a feel of subtleties of line fitting.