org.neuroph.contrib.model.errorestimation

Class Bootstrapping

java.lang.Object

org.neuroph.contrib.model.errorestimation.Bootstrapping

public class Bootstrapping
extends java.lang.Object

Error estimation method which uses sampling with repetition. Each sample contains N elements which forms so called E632 Bootstrap U odnosu na KFold nema nikakvu dodatnu logiku, samo drugi sampling i izbor datasetova tako da mogu kfold da spojim sa error estimation i da ovaj dobijam kroz setere ili parametrizaciju Vidi konacni kakva j erazlika izmedju ova dva ako je samo smpling metod, onda se moze koristiti korsvalifacija a drugim smpling metodom moze da se izvuce ErrorEstimation method kao osnovna klasa Boostraraping koristi sabsamples ne subsets http://www.faqs.org/faqs/ai-faq/neural-nets/part3/section-12.html mozes da koristis 632+ botstraping 50 - 2000 subsamples mogao bih da napravim sve to sa jednom klasom; KFold samo sa drugim sampling algoritmom

Constructor Summary

Constructors

Constructor and Description
Bootstrapping(int numberOfSamples) Default constructor for creating BootstrapEstimationMethod error estimation

Method Summary

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Bootstrapping

public Bootstrapping(int numberOfSamples)

Default constructor for creating BootstrapEstimationMethod error estimation

Parameters:

numberOfSamples - defines number of bootstrap samples