Developing Agent Based Modeling For Logic Programming And Reverse Analysis For Hopfield Network

• Main Author: Ng, Pei Fen
• Format: Thesis
• Language: English
• Published: 2013
• Subjects: QD1-999 Chemistry
• Online Access: http://eprints.usm.my/43864/1/Ng%20Pei%20Fen24.pdf; http://eprints.usm.my/43864/

For higher-order programming, higher order network architecture is necessary as high order neural networks have faster convergence rate, greater storage capacity, stronger approximation property, and higher fault tolerance than lower-order neural networks. So, higher order Hopfield network is brought into this thesis by using logic programming and reverse analysis in Hopfield network. The goal of performing logic programming based on the energy minimization scheme is to achieve the best global minimum. However, there is no guarantee to find the best minimum in the network. Thus, Boltzmann Machines and Hyperbolic Tangent activation function are being introduced to overcome this problem. To choose the best and efficient method to obtain the global minima among Wan Abdullah method (use McCulloch-Pitts updating rule in Hopfield net), Boltzmann machine and Hyperbolic Tangent activation functions, a comparison table will be created in this thesis. To carry out such work, agent based modeling (ABM) is created. NetLogo as the platform to carry out logic programming and reverse analysis. ABM can allow rapid development of models, easy addition of features and a user friendly handling and coding. In logic programming systems, not only the result in terms of global minimum will be analyzed but in the aspect of hamming distance and central processing unit (CPU) times will also be carried out. In reverse analysis systems, the inherent relationships among the data can be learned by extracting common patterns that exist in data sets. The unknown and unexpected relation can be seek. As a result, real life cases will be carried out by using ABM to run computer simulation in this thesis.