The main wo出 ofthls thesis Is to design and realize a Plant DiseaseDå«å¨osls Emert Sxstem based on Neural Nemork(PDDES NNNN)andmodemï¼dialing connectionæ¤ughthe research ofexpert systems basedmflclal neural network and that of remote senal communreatlonbetween PDA and PC.The details ofour Work are: 1.1讪1* *S,a*e娜rBPmuralmM0rh0dell*Sed,andforthe consideration oflts slow conveä¾£ent speed,one of*theeration two mainlåˆ s of*the basic BP algonthmhm,a new BP algonthln namedselfadå©tlve step factor BP algontå’–(SSFBP)has been put forward.The main idea of*the algonå’– Is:Change the value ofstep factor,l.e.theVç Œe OflCSYfllllgä»°e SCCOfdlllgtothC SignS &fldTC18tlVC SIZC OfthCMo latestwdlents as leamingprogresses,andå® to keep the currentStå°,l.C.thC SIZC OfWClght lflCTCffiCflt,SS idä¾£e SS pOSSlblC Wltholltovershooting so thatselFad地tatlon ofleaming rate can be reallied andMtcoM鸽t speedcanbe achieved.The algonth Is high eå¾ctlve.and easy to realize wlthoutneedingto choose any cntlcal parameters.ptthe same time,forthe consideration of*theeration probability oflocalCOllVå—mC* th OthCT Ofth tWO i81* 0åˆ S Of*Cçš„SIC BP &lgOTIå±±i,slmulatedannealingtheoryls In眈ducedto Integratethe sensltlvltyofthe micro s饲cture of*theï¼Šï¼ new BP algorithm with the macrodæ¤åœ°onality ofslmlated annealing algonth.So a combination of*theationtWO &lgoä¼½S,fi8ITICd SASSFBP slgofl讪,h肪 bCCflpfCSCfltCd lfl thisthesis.The main principle of*theple algonth Is:on one hand,t帅ugh thescWdwtlve step factorBP algontM,ltaccelerates thelteratlvepacesto cause fast convefgence In the negative gTadlent direction With thenumber ofuseless searches being greatly reå±±ced,and on the other hand,å’–ugh simulated annealing algonå±±m,It can.escaPe from local minimaso as to conveä¾£e to global minimum as much as possible.By theä»°etltlve alternation ofthetwo algonå’–s,the global minimum can beachieved inally. The results of our eXpenmentatlon,reflect that the two main lawsof*the basic BP algorith can be overcome to a large degree by theSASSFBP algont帅basedon combination of*theation above刚o algorithms. 2.AcomunlcatlonprotocolbasedonRSï¼232 asyncå€onous senalcomunlcatlongortandcormectlon establlshedbymodemdlaling,whichls suä¸å±±lefordå’–rismlsslonbetweenPDAandPC,has beendeslå³ed m this thesis.nd accoå·¾lug to this protocol,a remote senalcommunication systembetween PDAandPCbasedon Interruptï¼recelving mode,has been reallzednot only to solve the problem of*themremotedå’–trailsmlsslonbetween PDAandPC,butalso to conseectthebase forP以's urure"use ofaccessing InformationåŠ m expert systemsonPC servers and other Information from the Internet. 3·Basedonthe neural netwoå·¾'s learning algonth广一SASSFBPalgonthlnandthe rçš¿ote senalcoåˆ unlcatlon system,the yooESNNNNisdeslgnedandlmplememedlnthls thesis.The system overcomes thelåˆ s ofå’–dltlonal耶ertsystems such as邵atdlfficulty In knowledgeacålsltlon and low speedofreasoning to some extent,and has goodprospects Inthefieldofplå’–protection from diseases....
|