Global dynamics in neuro symbolic integration using energy minimization in mean field theory
Logic program and neural networks are two important aspects in artificial intelligence. This thesis is part of an endeavour towards neural networks and logic programming integration. The goal in performing logic programming based on the energy minimization scheme is to achieve the best ratio of glo...
Saved in:
| 格式: | Thesis |
|---|---|
| 语言: | English |
| 主题: | |
| 在线阅读: | http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/77974/1/Page%201-24.pdf http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/77974/2/Full%20text.pdf http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/77974/4/Muraly.pdf |
| 标签: |
添加标签
没有标签, 成为第一个标记此记录!
|
| 总结: | Logic program and neural networks are two important aspects in artificial intelligence. This thesis is part of an endeavour towards neural networks and logic programming integration. The goal in performing logic programming based on the energy minimization
scheme is to achieve the best ratio of global minimum. However, there is no guarantee to find the best minimum in the network. To achieve this, a learning algorithm based on the Boltzmann Machine (BM) concept and Hyperbolic Tangent Activation Function (HTAF) was derived to accelerate the performance of doing logic programming in Hopfield Neural Network (HNN) by using Mean Field Theory (MFT). Logic programming for lower order (up to third order clauses) and higher order clauses (up to eight order clauses) have been developed for MFT. The performance of this method is compared with the existing methods of doing logic programming in HNN (BM and HTAF). The global minima ratio, hamming distances and computational time were used to measure the effectiveness of the proposed method. Then, Agent Based Models (ABM) were developed by using Netlogo. ABM can allow rapid development of models, easy addition of features and a user-friendly handling and coding. Later the developed models are tested by using real life and simulated data sets. The simulation results obtain agreed with the proposed learning algorithm. The performance of doing logic programming using MFT proved to be better than the BM and HTAF. |
|---|