搜索
200210SHUILIXUEBAO10:20010703:(1965-),,,,,:05599350(2002)10011905BP黄胜伟1,董曼玲1(1,,271018):BP,(ABPM),,,,BP,ABPM,,:;BP;(ABPM);;:P6417;TP3016:A,70,,[1]30,,()(),,,,,,,[2],,,,,,,[2],,BP,7,BP(ABPM),1BP11BP(ErrorBackPropagationBP),BP2,,!119!∀∀∀4()[2],,BP,,,,E,E:E=12##(tjp-Ojp)2=#EP(1):E,PP,j,tjp,Ojp,(tjp-Ojp)2jP12BPBP[3]:W(k+1)=W(k)+D(k)(2):W(k);;D(k)=-E/W(k)kBP,:(1),,;(2),;(3),,13BP(ABPM)BP,BP,(ABPM)BPBP,E/w(k),,,;,,,,,,W(k+1)=W(k)+(k)[1-!D(k)+!D(k-1)](3):(k)=2(k-1);=sign[D(k)D(k-1)]:(k)k,(k-1)!,,,(3),,2ABPMABPM,,,,,21n,,7:(DO)(BOD5)(CN-)(Hg)(As)(Cr+6),,757,7,7m(),:!120!Xk=(x1,x2,∃,x7),k=1,2,∃,m(4):m22,,,,BP,,,[4],BPABPM,501Matlabrand(),()000011(INF)468101215BP8700INF6860INF36002650026802100022001620015808180ABPM103INF13855899339828531202182114140236811,8,,(1TNF);8~12,,,8~1223,7105,124RBF,1(GHZB1!1999),RBF,11(Hg)(As)%90200001000500000500500110000&6030000200500000500500501000∋50400005020000100500500100(30600010200010100500010)201000102000101010000125,,:x∗P1i=xP1ixi,max+xi,min,(i=1,2,∃,n)(5)t∗P1i=tP1jtj,max+tj,min,(j=1,2,∃,n)(6)(5)(6):xP1i,tP1jP1ij:x∗P1i,t∗P1jP1ij;xi,max,xi,mini;tj,max,tj,minj!121!(5)(6)[0,1]0139,1991~1993,221991~199393/(mg/l)/(mg/l)/(mg/l)/(mg/l)(Hg)/(mg/l)(As)/(mg/l)(mg/l)ABPMABPMBP13110000000000100021088-009900040000-0004%%%127150000000000011087-009700030000-0005%%%8029400080000001000080002-00020001-00020998)))1241500000000001087-009700040000-0005%%%8490001600000090012000002-0036022608050077(((60375005900180000900160001-00020002000100020998)))9278006300120000440076000018-0131110200070001∋∋∋80114003900480000380092000300040000000110530048(()125080000000001001088-009900040000-0005%%%2ABPM31ABPM!=03,(1)=001,ABPM,00001ABPMBP,,22ABPME;3ABPM(k)4BPE3ABPM(k)4BP323ABPMBP,,72%,1&,2∋,2(,∋;71%,5&,l),(!122!BP:(1)ABPM,BP,24,APBM,BP(2)ABPM,,,,,4ABPM,,,ABPMBP,,,,:[1].[M].::1988.[2],.[M].:,2000.[3],,.:[J].,1996,1(23):76-79.[4],.[J].,1999,1:60-62.[5],,.MATLAB[M].:,2000.ApplicationofadaptivevariablestepsizeBPnetworktoevaluatewaterqualityHUANGShengwei1,DONGManling1(1.ShangdongAgricultureUniversity,Tai+an271018,China)Abstract:Thebackpropagation(BP)networkistrainedusingadaptivevariablestepsizealgorithmtoovercometheshortageofconventionalBPNetwork.BasedontheobservationdatatheANNmodelforwaterqualityevaluationisestablished.Sevenpollutionindicesareadoptedastrainingstylebooktotrainthenetwork.Thetrainednetworkcanbeusedtoevaluatethewaterquality.ThecontrastcomparisonbetweentheconventionalBPnetworkandproposednetworkshowsthatthelaterpossessestheadvantageofhighaccuracyandhighspeedforconvergence.Keywords:artificialneuralnetwork;BPnetwork;adaptivevariablestepsizealgorithm!123!