Thông tin tài liệu
Ky ylu HQi thao ICT.rda'06 Proceedings of lCT.rda'06 Hanoi May VE M O T PHirOfNG PHAP XAC DINH CAU TRUC MANG NC NHAN TAO DUNG CHO L P CAM BIEN DO LlTOfNG On an Approach in Defining Artificial Neural Network Structi Using to a Class of Measure Sensors Nguyen Quang Hoan, Cao Minh Quyen, Nguyen Tai Tuyen Tom tdt Bdi bdo trinh bdy phuang phdp xdc dinh cdu true mgng na ron nhdn tgo dimg cho lap cd( cdm biin luang Bdi bdo trinh bdy djnh ly cho thdy mgng na ron truyin thdng mot l&f dn dii de dnh xg hdm md td dgc luyin aia cdm biin luang bdt ky Dinh ly Id ca ^ Jt chgn cdu tnic mgng na ron fMNN^ niian lao img dung cdc cdm biin ludng Tir khod: mgng na ron (MNN); cdm biin luang: hdm md td dgc tuyin Abstract This paper presents an approach in defining artificial neural network structures using to a class of measure sensors The theorem is proposed and indicated that, the feedfon\'ard neural networks with one hidden layer are enough mapping from sensor measure patterns to measure characteristic function This theorem is basic fundamentals to choose neural networks structures using in measure sensor techniques Keywords: artificial neural networks; measure sensor; characteristic function De khdc phuc ban ehl nay, mdt sd n ciiru khac da ung dung tinh toan no ron cl Trong ludng, cau true cija cac phuong toan ludng [3, 4] Mgt nhung d^ tif n va hf thdng quyet djnh cac ddc tinh ky' la ghep MNN vdi chuyin ddi so cip dl thuat CO ban nhu: dg nh|y, dg chinh xac, cao cdc ddc tinh ky thuat cho cam bie ddi lam vifc v.v Nhieu cdng trinh nghien ludng ciiru nhim nang cao chat lugng cac ddc tinh ky Khi giai bai toan neu tren, mdt vi thuat ciia cam bien [3, 4, 6, 11] Cac nghien xuat hifn la co sd dl lira chgn MNN Ban cuu dd hau hit tap trung theo hai hudng ciia bai toan la chgn MNN dl anh xa ban Hudng thir nhdt: Nghien ciiru cdng nghf tuyen nhieu biln, nhirng ddc tuyln lam che tao de nang cao ddc tinh ky thuat cua ciia cam bien ludng chuyen ddi so cap Day la hudng nghien ciiu Trong ly thuylt MNN cd dinh ly cd tinh kinh dien va khd phat trifn doi vdi Kolmogorov phat bilu nhu sau "Cdu true dieu kifn Vift nam MNN du di xdp xi mgt hdm liin tyc nhieu I Huang thir liai: Ung dung khoa hgc hifn phi tuyin bdt ky phdi cd tdi thiiu ldp c dai de nang cao ddc tinh ky thuat cua cam djnh ly khdng chi ro ham truyin cua cac bien ludng Tir cudi the ky 20, nhilu iimg ron ldp in cin dugc chgn [8, 10, 121 dung ky thuat Vi xiir ly vao ludng da cho Bai toan cam biln la bai toan hep nhieu cdng Tuy nhien, vifc ung dyng ve gidi han, la anh x^ mgt ham nh ky thuat Vi xiir ly de \u ly thdng tin cd thdi bien, phi tuyen nhung don tri va hiru han gian th^c hifn cac lfnh anh hudng tdi toe dg I DAT VAN DE yAyj« Hfli thto ICT.rda'06 Djnh nghia, ky hifu vaphdt bieu djnh ly D9c tuyln lam vifc cua mdt cam bien ludng la mgt ham rdi rac nhieu biln, phi tuyln nhung don trj va bu-u han mdt khoang [a.b] Ddc tuyen cd the dugc cho bdi tap hgp sd lifu dudi d^ng y^(x^), k=l,2 n MNN cd r diu vao, mgt Idp an vdi N no ron va mgt diu dugc ky hifu bdng cdng thurc cau true r-N-1 B6 DE VA DJNH LY B6de (MNN 1-2-1 dnh xg I diem citayifxi)): MNN vdi ciu true r-2-1 du 6ik dnh xa mgt dilm bit ky cua yk(xk) Chiimg minh: Mang mang no ron vdi ciu true 1-2-1 {hinh 1) anh xa dugc dilm (x,,y,) bit ky 1) Trudng hgp anh x^ mgt dilm X| tu diu vao sang khdng gian diu az = yi Proceedings of ICT.rda'06 Hanoi May 20-21, 2006 Ham truyen no ron Idp an chon la ban bude nhay {Hardlims) Md hinh eiia eac n( ron cap thur nhit ldp in dugc dien ta: a',= Hardlims[w'i.(x,+ 8x)] a\= Hardlims[w'2 (X|- 8x)] Gia sir yi > (1) (2) w'l
Ngày đăng: 08/12/2022, 21:06
Xem thêm:
