r: Choose number of KNNs

In rknn: Random KNN Classification and Regression

Description

Choose number of KNNs

Usage

r(p, m = floor(sqrt(p)), eta = 0.99, nu = 20, rmax = p, nsim = 1000,
    lambda = 0.01, method = c("binomial", "poisson", "nu", 
    "geo.independent", "geo.sim", "geo.dependent", "lambda"))

Arguments

p	Total number of available features.
m	Number of features to be drawn by each KNN.
eta	Coverage Probability.
nu	mean mutiplicity of a feature
rmax	number of series terms for independent geometric approximation
nsim	number of simulations for geometric simulation.
lambda	mean number of silient features.
method	one of binomial, poisson, nu, geo.independent, geo.sim, geo.dependent, lambda

Details

Method binomial and poisson are approximation method for a given eta value.

Method nu computes r for a given nu.

Method geo.independent, geo.sim and geo.dependent compute r using geomerical random variables z until eachh is at least drawn once. The difference is that geo.independent ignores dependency,geo.sim is a simulation method and geo.independent is an exact method using gmp package.

Method lambda computes r for a given lambda.

Value

An integer.

Author(s)

Shengqiao Li<[e-mail]>

Examples

  r(100, 10);

Example output

Loading required package: gmp

Attaching package: 'gmp'

The following objects are masked from 'package:base':

    %*%, apply, crossprod, matrix, tcrossprod

[1] 88

rknn documentation built on May 2, 2019, 12:35 p.m.