Class fit_bayes (o2scl)

O2scl : Class List

template<class fit_func_t = fit_funct, class multi_func_t = uniform_prior<>, class vec_t = boost::numeric::ublas::vector<double>>
class fit_bayes

Fit a function to data using Bayesian methods.

This class is experimental.

This class uses Markov Chain Monte Carlo (MCMC) and marginal estimation to give a probability distribution for parameters in a fit.

Idea for Future:

Also make weight_fun() an object of type multi_func_t?

Offer two ways to do the evidence: direct MC or the interpolation method from SLB13

Build upon gen_fit_funct instead of fit_funct?

Public Types

typedef boost::numeric::ublas::vector<double> ubvector

typedef boost::numeric::ublas::matrix<double> ubmatrix

typedef boost::numeric::ublas::vector<int> ubvector_int

Public Functions

inline fit_bayes()

inline virtual void evidence(size_t ndat, vec_t &xdat, vec_t &ydat, vec_t &yerr, size_t npar, vec_t &plo2, vec_t &phi2, multi_func_t &prior_fun, double &evi, double &err)

Compute the evidence.

inline virtual double weight_fun(size_t ndat, const vec_t &xdat, const vec_t &ydat, const vec_t &yerr, size_t npar, const vec_t &par)

The weight function (based on a \( \chi^2 \) distribution)

inline virtual int fit(size_t ndat, vec_t &xdat, vec_t &ydat, vec_t &yerr, size_t npar, vec_t &plo2, vec_t &pmax, vec_t &phi2, vec_t &plo_err, vec_t &pmax_err, vec_t &phi_err, fit_func_t &fitfun, multi_func_t &prior_fun)

Fit ndat data points in xdat and ydat with errors yerr to function fitfun with npar parameters.

The initial values of the parameters should be specified in par.

inline virtual int fit_hist(size_t ndat, vec_t &xdat, vec_t &ydat, vec_t &yerr, size_t npar, vec_t &plo2, vec_t &phi2, std::vector<hist> &par_hist, fit_func_t &fitfun, multi_func_t &prior_fun)

Desc.

For each measurement block, this function collects the data for all the parameters into 1d histogram objects. Then, at the end of the block, the histogram information is added to a hist object for each parameter.

Public Members

size_t n_warm_up

Number of warmup iterations (default 100)

size_t n_iter

Number of total iterations (default 1000)

size_t hsize

Histogram size (default 20)

size_t nmeas

Number of measurements (default 20)

multi_func_t *pri

Prior distribution.

o2scl::rng gr

Random number generator.

mcarlo_vegas def_inte

Default Monte Carlo integrator.

Protected Functions

inline virtual double integrand(size_t npar, const vec_t &par)

The integrand for the evidence.

Protected Attributes

fit_func_t *ff

User-specified function.

size_t lndat

Number of data points.

vec_t *lxdat

X-values.

vec_t *lydat

Y-values.

vec_t *lyerr

Y-errors.