Share
تحسين التمييز للشبكات العصبية المنعدمة الاوزان == Recognition Enhancement of Weightless Neural Network
Author name:
عمار علي مصطفى
Supervisor name:
محمود حمزة المفرجي
General topic:
Electrical, Electronic and Communications Engineering
Specific topic:
Electrical Engineering
Degree:
Master
University:
University of Technology - Electrical Engineering Department - Electrical Engineering Branch
Language:
English
University location:
Baghdad
Key words:
- Pattern Recognition
- RAM - based Weightless Neural Network
First pages:
34T513 - p.pdf
Abstract:
الفعل الذكي للشبكات المتعلمة في هذا البحث يركز على الشبكات المتعلمة في منظومة تمييز الانماط المعتمدة على الذاكرة. ان الفعل الذكي في هذه المنظومات يكمن في القابلية على تمييز الانماط التي لم تتدرب عليها مسبقا وذلك يتم بالاعتماد على ظاهرة التعميم. الهدف من ا | The intelligent action of the learning network in this research concentrates on the learning network in the pattern recognition systems that are based on the Random Access Memory. The intelligence action in the pattern recognition system is the capability to recognize other patterns that were not learned before in the system depending on the generalization. The objective of the research is oriented to improve the performance of the learning network in pattern recognition system. A pattern recognition system that based on Random Access Memories is developed and evaluated for classifying non - deterministic data with particular reference to unconstrained handwritten Arabic numerals and traffic signs patterns.Various techniques are presented which allow the system to be optimized, giving an increase in the performance and confidence of the results.The results obtained showed the improvements of n - tuple and VG - RAM WNN systems as learning network. Techniques used to enhance the performance are investigated. The averaging of the training patterns for colored images is the proposed system. This technique gives improvements to the system performance in term of required storage size and recognition speed.
