SavedAdd your journal

Journals / Anadolu Üniversitesi Bilim ve Teknoloji Dergisi :A-Uygulamalı Bilimler ve Mühendislik / 2002 / Cilt: 3 - Sayı: 1

A multi-scale approach for selecting the optimized activiation function in multi-layer artifical neural networks

Çok katmanlı yapay sinir ağlarında en iyi etkinleştirme fonksiyonu seçimi için çok ölçekli bir yaklaşım

İbrahim TÜRKOĞLU
Ahmet ARSLAN
Pages
137–142
Publication Date
2022-10-11
DOI
—
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

Hence, by shortening the selection process of the activation function, which takes long time and requires experience, the desired high speed and accuracy are obtained in the pattern recognition.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:In this study, a new algorithm using the distributed system structure is developed in order to improve the training of the artificial neural network classifiers used in pattern recognition.
📊
Key Findings & Data:With this algorithm, selection of the most suitable type of the activation functions for the artificial neural networks can be realized by trying with a multi-scale approach usage all the possibilities in the distributed system within a short time.
💡
Conclusion & Impact:Hence, by shortening the selection process of the activation function, which takes long time and requires experience, the desired high speed and accuracy are obtained in the pattern recognition.
🔬Design:Research Article
📍Scope:Comprehensive Sample
⚙️Variables:Domain Parameters

Abstract

In this study, a new algorithm using the distributed system structure is developed in order to improve the training of the artificial neural network classifiers used in pattern recognition. With this algorithm, selection of the most suitable type of the activation functions for the artificial neural networks can be realized by trying with a multi-scale approach usage all the possibilities in the distributed system within a short time. The algorithm is tested for the pattern recognition of the radar receiver signals by using the multi-layer neural networks with the back propagation-learning rule. Hence, by shortening the selection process of the activation function, which takes long time and requires experience, the desired high speed and accuracy are obtained in the pattern recognition.

Özet

Bu çalışmada, örüntü tanımda kullanılan çok katmanlı yapay sinir ağı sınıflandırıcılarının eğitimini iyileştirmek için dağıtık sistem temelli yeni bir algoritma geliştirilmiştir. Bu algoritma ile yapay sinir ağı için en uygun etkinleştirme fonksiyon türü seçimi, dağıtık sistemde tüm olasılıklar çok ölçekli bir yaklaşımla denenerek kısa sürede gerçekleştirilebilmektedir. Algoritma, hata geri yayma öğrenme kuralını kullanan çok katmanlı yapay sinir ağlarına uygulanarak radar alıcısı sinyal örüntülerinin tanımlanması ile test edilmiştir. Böylece uzun zaman alan ve tecrübe gerektiren etkinleştirme fonksiyonu seçim süreci kısaltılarak, örüntü tanımada istenen yüksek hız ve doğruluk sağlanmıştır.

🔖

Cite This Article

TÜRKOĞLU, İ., ARSLAN, A. (2022). A multi-scale approach for selecting the optimized activiation function in multi-layer artifical neural networks. Anadolu Üniversitesi Bilim ve Teknoloji Dergisi :A-Uygulamalı Bilimler ve Mühendislik, Cilt: 3 - Sayı: 1, 137–142. https://ajindex.com/articles/438514

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (11)

  1. Alba, E. veTroya, J. M. (2000). Analyzing synchronous and asynchronous parallel distributed genetic algorithms. Future Generation Computer Systems, 17(4), 451-465.
  2. Bishop, C.M. (19%). Neural Networks for Pattern Recognition. Clarendon Press,
  3. Oxford. Duda, R.Ü. ve Harı, PH. (1989). Pattern Classification and Scene Analysis. Stanford Research Institute, New York. Gökbulut, M. (1998). Fırçasız Doğru Akım Motorlarının Yapay Sinir Ağları ile Uyarlamak Denetimi. Krciyes Üniversitesi Fen Bilimleri Enstitüsü Doktora Tezi. Haykin, S. (1994).
  4. Neural networks, A Comprehensive Foundation. Macmillan College Publishing Company Inc., New York. Hunt, K.J., Sbarbaro, D. ve Zbikowski, R. (1992). Neural Networks for Control Systems-A Survey. Automatica. 28(6), 1083-1112.
  5. Lab-Volt (1993). Radar Eğitim Seti. YÖK/Dünya Bankası II. Endüstriyel Eğitim Projesi. Lin, C.T. ve Lee, C.S.G. (1996). Neural Fuzzy Systems. Prentice-Hall. Narayan, S. (1997). The Generalized Sigmoid Activation Function Competitive Supervised Learning. Elsevier Science Inc. Information Sciences, 99 (1-2), 69-82.
  6. Narendra, K.S. ve Parthasarathy, K. (1990). Identification and Control of Dynamical Systems Using Neural Networks. IEEE Trans, on Neural Networks, 1(1), 4-27.
  7. Pao, Y.H. (1989). Adaptive Pattern Recognition and Neural Networks. Addison-Wesley. Türkoğlu, î. ve Arslan, A. (2001). Optimisation of the Performance of Neural Network Based Pattern Recognition Classfiers with Distributed Systems. IEEE Computer Society, 2001 International Conference on Parallel and Distributed Systems (ICPADS'2001), ss.379-382,
  8. Kyong Ju, Kore. Wittie, L.D. (1991). Computer Networks and Distributed Systems. IEEE Trans, on Computers, 24(9), 67-76.
  9. Young, T.Y. ve Fu, K.S. (1986). Handbook of Pattern Recognition and Image Processing. Academic Press.
  10. Zhu, W, Liang, T.Y. ve Shieh, C.K. (1999). A Hopfield Neural Network Based Task Mapping Method. Elsevier Computer Communications, 22, 1068-1079.
  11. Zurada, MJ. (1992). Introduction to Artificial Neural Systems. West Publishing Company Inc., New York.