Research And Application Of Extended Adaptive Spiking Neural P System

With the development of computer technology and biological science technology,nature provides a rich source of ideas for human beings to build new computational models.As we all know,the human brain is the most powerful biological computer in nature.The human brain contains hundreds of billions of neurons.Each neuron is not an independent individual.They are highly interconnected through synapses,and thus constitute complex neural network structure.By simulating the connection mode and information processing mode of neurons in human brain,a series of neural computing models,such as artificial neural network and spiking neural P system,have been put forward successively.The main feature of the artificial neural network is that it can independently learn knowledge in the environment and continuously improve its performance.It has made great achievements in neural computing theory and production applications.Spiking neural P system has been widely used in biological sciences,computer science,image recognition and other fields with its powerful computing power.As a computational model inspired by the nervous system of the human brain,the spiking neural P system has similar topological structures to the artificial neural network.Therefore,it is of great theoretical significance and practical value to study the "cognition and learning" ability of the spiking neural P system.From a theoretical perspective,(1)this thesis firstly proposes a new type of spiking neural P system,namely the extended adaptive spiking neural P system,which improves the content of neurons,the structure of the system,and the rule set on synapses.By simulating the register machine,it has been proved statically that the new P system constructed is Turing-computable in both the generation mode and the acceptance mode.(2)In order to make the spiking neural P system have completely adaptive ability,and make it more in line with biological reality,this thesis specifies the learning function in the extended adaptive spiking neural P system,and introduces the dynamic learning rules in the neural network into the extended adaptive spiking neural P system.The extended adaptive spiking neural P system based on Hebb learning rules and the extended adaptive spiking neural P system based on SOM competitive learning rules areconstructed respectively.This provides a new idea for the study of the spiking neural P system.From the perspective of application research,the extended adaptive spiking neural P system based on Hebb learning rules is firstly applied to English character recognition in this thesis.The English character image is converted into binary information,and then converted into spiking information to be input into the system.Then,through the expansion of input block and the training of recognizing block,the corresponding output information is generated for judgment and recognition.Through specific simulation experiments,it is verified that the average accuracy of the character recognition under the noise-free system can get 99.4%,and even at high noise levels,the average accuracy of the system recognition is more than 70%.The feasibility of this system in the field of English character recognition is proved,which indicates that the spiking neural P system has potential application value in the field of image recognition.In addition,the constructed extended adaptive spiking neural P system based on SOM competitive learning rules is applied to the clustering problem.The real number information in the specific data set is input into the system,and then through the information preprocessing of input block and the clustering operation of clustering block,the information of different data sets is classified.The UCI database is input into the system,and it is proved that the extended adaptive spiking neural P system based on SOM competitive learning rules performs well in clustering applications,which also effectively broadens the research scope of the spiking neural P system.