内容紹介

Linear Discriminant Analysis (LDA) has been researched for the computational intelligence on network security applications. In the context of present and future network communication and security, this talk introduces a series of LDA new developments, where an LDA model is enabled to be learned either in one batch session, or incrementally by Incremental LDA (ILDA), or through LDA eigenspace merging (LDA Merging); For parallel computing, a global LDA can be learned cooperatively by a group of regional LDA agents with their knowledge shared each other while learning (mILDA); In a special case, a created LDA can be renovated by splitting LDA eigenspace with a minimum processing on the raw data instance (LDA Splitting); For even higher performance computing, LDA can be set to be learned only on fewer selected curiosity instances (cILDA), or by multiple cILDA agents in a competitive and cooperative learning manner (mcILDA).

詳細情報

刊年

2010

形態

電子化映像資料(1時間06分56秒)

シリーズ名

情報科学研究科・ゼミナール講演 ; 平成22年度

注記

Auckland University of Technology

講演日: 平成22年12月22日

講演場所: 情報科学研究科大講義室L1

標題言語

英語 (eng)

本文言語

英語 (eng)

著者情報

Pang, Shaoning