Loading...

Presentation of a Processing Structure with Ability of Chaotic, Fuzzy and Neural Models

Esmaili Paeen Afrakoti, Iman | 2014

841 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 46044 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Bagheri Shouraki, Saeed
  7. Abstract:
  8. Research on how the human brain processes information leading to the creation of two major groups in the field of soft computing. The first group believes that information is being processed based on linguistic concepts and if-Then rules. Fuzzy logic is based on this idea and tries to avoid exact calculations in information processing tasks. Second group believes that the power of human brain in processing is because of a large network of neurons with small abilities. These studies led to the presentation of artificial neural networks algorithms. Spiking neural network is known as third generation of artificial neural networks and tries for processing information using a real model of brain system and the timing of electrical pulses. Also for more accurate modeling of neural network in human brain, complex mathematical equations are used for modeling a variety of simple and complex behaviors (chaos) of neurons. In this thesis a novel algorithm based on active learning method and ink drop spread operator as an efficient fuzzy algorithm is proposed which implements the fuzzy concepts using simple model neurons in spiking artificial neural networks. The target of this thesis is integrating the concepts of two introduced groups. In order to eliminate weaknesses and enhance the efficiency, new structure is provided which modeling task is done in higher dimensions. Result of thesis is presentation of a general model which can be implemented as second and third generations of artificial neural network based on fuzzy concepts with no need of optimization algorithms in training phase. Simulations results confirm the performance of the proposed algorithm in complex modeling applications such as prediction of chaotic behavior and pattern recognition. One of the major challenges of fuzzy systems and neural networks is their hardware implementation. Suitable hardware is designed based on memristor crossbar structures for each presented algorithm
  9. Keywords:
  10. Learning Algorithm ; Ink Drop Spread (IDS)Operator ; Memristor ; Spiking Neural Network ; Spiking Artificial Neural Network ; Chaotic Behavior

 Digital Object List

 Bookmark

...see more