TRIMMED AND WINSORIZED M- AND  ESTIMATORS, WITH APPLICATIONS TO ROBUST ESTUTION IN NEURAL NETWORK MODELS

منو

" TRIMMED AND WINSORIZED M- AND ESTIMATORS, WITH APPLICATIONS TO ROBUST ESTUTION IN NEURAL NETWORK MODELS "

Document Type	:	Latin Dissertation
Language of Document	:	English
Record Number	:	151505
Doc. No	:	ET23297
Main Entry	:	Zhenwu Chen
Title Proper	:	TRIMMED AND WINSORIZED M- AND ESTIMATORS, WITH APPLICATIONS TO ROBUST ESTUTION IN NEURAL NETWORK MODELS
Note	:	This document is digital این مدرک بصورت الکترونیکی می باشد
Abstract	:	Wmmed and Winsonzed M- aad Z-estimators are introduced as new a pproaches toward robust estimation. The consistency, convergence rate of n-i,and asymptotic normality of the estimators are established. It is shown thatmany robust estimators can be derived From the framework of trimmed andWinsonzed M- and estimators and thus be treated in a unified way. Theapproach also sheds light on the search for new robust estimators, especiallyin complicated models. Examples of some well-known robust estimators arediscussed and some new robust estimators are derived. Application to robustestimation in neural network models is also dimmed.....-....,....-...,..tested for theQ1 PC1 bus cardBoth these projects mere sofixare des elopment efforts tonards contributing to dlfferentaspects of Roboucs and lZ1echatronics projects m the Controls and Roboucs Group..
Subject	:	Electericl tess
	:	برق
electronic file name	:	TL46534.pdf
Title and statement of responsibility and	:	TRIMMED AND WINSORIZED M- AND ESTIMATORS, WITH APPLICATIONS TO ROBUST ESTUTION IN NEURAL NETWORK MODELS [Thesis]