Loading...

Alm Improvement Based On New Fuzzy Operator With Memristor Implementation Capability

Haghzad Klidbary, Sajad | 2018

590 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 51541 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Bagheri Shouraki, Saeed
  7. Abstract:
  8. Designing artificial intelligence based arithmetic machines that can intelligently perform human-like task has attracted considerable interest among researchers. The ever-increasing advances in soft-computing algorithms require appropriate hardware platforms for such algorithms. One of the most important problems with these algorithms and their hardware implementation structures is the discrepancy between the hardware and the nature of the problem. It can be argued that paying attention to hardware implementation does not necessarily guarantee an optimal implementation of these algorithms. Most of the proposed hardware implementations have very small resemblance to the biological systems (e.g. human brain) in performing those computations; therefore, this discrepancy reduces the efficiency of the system and demand significant hardware resources.Inspiration from the nature has always been in the attention of researchers; however, the implementation of brain-like structures has many challenges. The brain has a very large number of neurons and synapses and regarding the volume of the brain, it can be concluded that the density of neurons is very high. So far, many computational structures have been built on the basis of CMOS technology, but due to the fact that this technology encounters some problems in the nanoscale, it is not a good option for implementation of Bio-Inspired systems and reaching the high volume of processing units. Therefore, the main reasons that researchers and engineers have not achieved satisfactory development in this field are first technological challenges and second the lack of sufficient understanding of nervous systems and computational biological system.In recent years with realization of Memristor, a new horizon for researchers has been opened up and we have become closer than ever to a commutating platform with considerable resemblance to a biological nervous system. However, it seems that in addition to appropriate processing units, the algorithms should employ less complex computations; so that they can take an effective step in the artificial intelligence path. Active Learning Method (ALM) is one of the most suitable methods that has shown good results in various applications. This thesis is based on two unique property of ALM, fuzzification of inputs suing ink drop spreading operation and analyzing a complex system with multiple simpler sub-systems. Therefore, the aim of this thesis is to first improve ALM and second optimize its hardware implementation. In the following, an efficient algorithm without the need for optimization based on the ink drop spreading concepts is presented and finally, a fuzzy operator is presented.The most important contributions of this thesis are 1. introducing adaption capability to the ink spreading radius, 2. introducing an alternative algorithm in the IDS operator, and 3. proposing low-volume hardware implement of ALM on memristor-crossbar and FPGA platforms. Moreover, using the idea of spreading ink drops, a new algorithm with lower computational cost (without many parameters, complex calculations and optimization algorithms), lower volume and higher speed in modelling and classification of low as well as high dimensional data is proposed. Finally, a new computational paradigm that both the operands and operators are fuzzy is presented. In all stages of this thesis, low computational complexity, low-volume hardware with implementation capability, as well as hardware with learning capability have been considered
  9. Keywords:
  10. Active Learning ; Fuzzy Modeling ; Ink Drop Spread (IDS)Operator ; Fuzzy Inference System ; Memristor ; Field Programmable Gate Array (FPGA) ; Spiking Neural Network ; Fuzzy Operator

 Digital Object List

 Bookmark