Geri Dön

Development of conic section function neural networks in software and analogue hardware

Başlık çevirisi mevcut değil.

  1. Tez No: 400512
  2. Yazar: TÜLAY YILDIRIM
  3. Danışmanlar: DR. JOHN S. MARSLAND
  4. Tez Türü: Doktora
  5. Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Matematik, Bilim ve Teknoloji, Computer Engineering and Computer Science and Control, Mathematics, Science and Technology
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 1997
  8. Dil: İngilizce
  9. Üniversite: Unıversıty Of Lıverpool
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Yapay Zeka Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: Belirtilmemiş.

Özet

.

Özet (Çeviri)

The ability of the biological brain to perform very complex tasks has inspired scientists to study the field of neurai computation and to try to implement artificial neural systems. in recent years, many neural algorithms and several neural hardware systems have been proposed. This thesis is concerned with the use of conic section functions which contains RBF (Radial Basİs Functİon) and MLP (Multilayer Perceptron) networks. The work is concentrated in two areas: The implementation of learning algorithms for the Conic Section Functİon Neural (CSFN) netvvork and the implementation of the CSFN in analogue hardware. A new training algorithm composed of a propagation rule which contains MLP and RBF parts to improve the performance of back propagation is proposed. The network using this propagation rule is known as a Conic Section Functİon Network. This network converts the öpen decision boundaries in an MLP to closed ones in an RBF. A training algorithm has been implemented in MATLAB. The performance of the proposed algorithm is compared with existing MLP and RBF algorithms. An analogue VLSI hardware design for a Conic Section Functİon Neural netvvork vvhich allows the use of RBF and MLP propagation rules on a single chip, depending on the data distribution of a given application, is proposed. CSFN contains hyperplane and hypersphere decision regions for MLP and RBF, respectively. A novel synapse and neuron circuİt for a CSFN has been designed in analogue hardware to compute both the dot product (vveighted sum) for MLP and the Euclidean distance between input vectors and centres for RBF. These two propagation rules are then aggregated to form a conic section function netvvork. The designed circuits vvere simulated using cdsSpice simulator in Cadence design package for different number of synapses and neurons. Two chips, a synaptic circuit for a CSFN and a Conic Section Functİon Neural netvvork, have been designed using Cadence design package vvith Mietec 2.4um technology. These chips have been fabricated and tested.

Benzer Tezler

  1. Genel amaçlı bir yapay sinir ağının karma bir donanımla gerçeklenmesi

    Mixed mode hardware design of a general purposed artificial neural network

    BURCU ERKMEN

    Doktora

    Türkçe

    Türkçe

    2007

    Elektrik ve Elektronik MühendisliğiYıldız Teknik Üniversitesi

    Elektronik ve Haberleşme Mühendisliği Ana Bilim Dalı

    PROF.DR. TÜLAY YILDIRIM

  2. Yapay sinir ağı ve bulanık mantık hibrid yapı ve algoritmalarının geliştirilmesi

    Development of artificial neural network-fuzzy logic hybrid structures and algorithms

    CANAN ŞENOL

    Doktora

    Türkçe

    Türkçe

    2010

    Elektrik ve Elektronik MühendisliğiYıldız Teknik Üniversitesi

    Elektronik ve Haberleşme Mühendisliği Ana Bilim Dalı

    PROF. DR. TÜLAY YILDIRIM

  3. Ovulasyon indüksiyonu tedavisinde folliküler gelişimin ultrasonografik takibi

    Başlık çevirisi yok

    MERİH BAYRAM

    Tıpta Uzmanlık

    Türkçe

    Türkçe

    1987

    Kadın Hastalıkları ve DoğumGazi Üniversitesi

    Kadın Hastalıkları ve Doğum Ana Bilim Dalı

    DOÇ. DR. MÜLAZIM YILDIRIM

  4. Türkiye'de ihracatın finansmanı

    Başlık çevirisi yok

    MEHMET BAHA KARAN

    Doktora

    Türkçe

    Türkçe

    1984

    EkonomiGazi Üniversitesi

    Muhasebe ve Finansman Ana Bilim Dalı