A self-generated neural network construction method

Abstract

A method for constructing a self-generated neural network, comprising the following steps: Step 1, adding a stimulus signal; Step 2, evaluating the neuron output strength, determining the connection direction of the neuron, continuously forming network connections, and finally generating an initial network; Step 3, Calculate the position and probability of connecting to the target neuron; step 4, judge whether the current network generation process is stopped, if so, then go to step 5, otherwise go to step 2 to continue; step 5, optimize the network connection through the optimization algorithm; Step 6, judge whether it is necessary to add stimulation, if not, then end, otherwise go to step 1; the network proposed by the present invention is a self-generated network, which can effectively reduce the influence of human subjective factors. Based on neural theory, it is possible to further explore the neural brain and realize real intelligence.

Description

technical field [0001] The invention belongs to the technical field of neural network calculation, relates to a software-oriented modeling method and a hardware-realized neural network self-generation method, in particular to a self-generated neural network construction method. Background technique [0002] With the development of computers today, both computing power and power consumption have been substantially improved in performance. But with the diversification of people's needs, the current computer structure exposes more and more problems. There are two types of information processing problems provided by nature: structured and unstructured. A structural problem means that it can be clearly and strictly described in mathematical language, and the realization algorithm of the problem can be formulated and mapped to a computer program, and then processed by computer instructions. Traditional von Neumann machines far outperform humans in solving such problems. Non-str...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the object of the present invention is to provide a method for constructing a self-generated neural network, which maps the characteristics of the biological neural model and the appropriate heuristic algorithm to the silicon-based circuit, so that the neural network can pass the model Self-generating and self-organizing, avoiding the shortcomings of artificial neural networks

Images