MonteCarlo: MonteCarlo simulation for optimization

Description Usage Arguments Value

View source: R/functions2eval.R

Description

Make a MonteCarlo simulation from initial value on each schemes' coeficients. Initial sample has a 'nsample'*10 population number for each parameter in scheme, determined by a LHS (Latin Hypercubic Sample) with frontiers +- 5 times the value of coeficient: $$ A_i = [a_i-5a_i;a_i+5a_i] $$ The MonteCarlo design follows a similar way in genetic algorithms, where a initial population of value for parameter produce a L_i serie that is evaluated with choosen estatistic (stats). Always two equal dimenson population are compared and choosen combinations of parameters that produce a 'stats' minor than a o.5 quantile. That way ensures that best combinations be conserved if its performance is better than a half of total combinations 'stats' value. Finally a coeficients combination with best performance prevales and is considered the optimization. Furthermore, a 'nsample', 'max_iter' and 'stats' will define a result.

Usage

1
MonteCarlo(data, E_fun, func, coefs, nsample, max_iter, stats)

Arguments

data

Data frame with all atmospherics variables

E_fun

Function for emissivity

func

Function from cloud cover calculation

coefs

Coeficients in E_fun function

nsample

Number of population individuous

max_iter

Maximun number of iterations

stats

Statistical index to be minimized

Value

Vector with best combination of parameters


roilanhv/longwaveR documentation built on Aug. 24, 2017, 3:02 a.m.