基于过拟合神经网络的混沌伪随机序列.pdfVIP

第9卷 第 11期 2009年6月 科 学 技 术 与 工 程 Vo1．9 No．1l June2009 1671—1819(2009)11-3101—04 ScienceTechnologyandEngineering ⑥ 2009 Sci．Tech．Engng~ 基于过拟合神经网络的混沌伪随机序列 罗松 江 丘水生 (华南理工大学电子与信息学院，广州510640) 摘 要 结合神经网络和混沌映射的特点，提出了一种基于过拟合BP神经网络的混沌伪随机序列产生方法。以logistic映射 和Henon映射为例，测试结果表明，该方法能克服有限精度效应对混沌系统的影响，改善混沌伪随机序列的性能，为获得高性 能的伪随机序列提供了一种新的思路。 关键词 过拟合 神经网络 伪随机序列 中图法分类号 TP183； 文献标志码 A Pseudorandom sequencehas alotofcharacters cisioneffectswhen theyare implemented incomput— similartowhitenoise，andhasbeenwidelyusedinse— er [ ． pseudorandom sequencehtatgetfrom low di— cure communication，cryptography，automatic control mensionchaoticsystem willbecomeshortperiodand andotherraeas．Cryptorgaphicprotocolsforelectronic cannotmeettherequired． paymentsystems，authentication，integrityorkeyman— A mehtodofproducingchaoticpseudorandom se— agementmayhaverandom components，whichrequire quencebasedonoverflttingneuralnetworkisproposed methodstoobtainingnumbers thatarerandom insome inhtispaper．Testresultsindicatethatitsperfomr ance sense，So，analyzing na d designing rna dom bitse— isbetterhtanthesequenceobtainedfrom simplyquan·· quencesofgoodquality，i．e．，ofgoodbehaviorinsta— tify the chaotic systems．Even iftraining timesand tisticalandnon—predictabilitytermshasbecomeahot trainingsamplesarethesame，thereceivedpseudoran— pmblem toresearchers[． dom seuqenceisdifferentwiht htedifferenceofinitial Ingeneral，chaoticsystemspossessseveralprop- weightofneural network．The characterismoresuit— ertieswhich makethem essential componentsin con— ableofrallkindsofapplication． structingpseudorandom sequence：(i)Randomness： chaoticsystemsgeneratelong—period，random-likecha— 1 Principle ofproducing c