整合神经网络的人工生命:一种脑回路研究方式

Neural Networks-Intergrated Artificial Life: a New Approach to Study Functional Neural Circuit

基于设计能够自动挑选出有特定功能的模拟神经回路这一设想,通过给虚拟生物体一个未连接的模拟小神经群,在与虚拟环境的相互作用下,研究其神经网络的拓扑结构的演化。具体做法是:让虚拟生物体的小神经群在最初视觉神经连接至中枢神经,中枢神经内部无相互连接,中枢神经与运动神经不相连接,运动神经元可以有一定概率的不通过中枢神经而随机激发。然后在虚拟环境中投入食物,人工生物体运动神经元自动激发造成食物与生物体相对位移的改变,进而导致Hebb学习,造成神经网络拓扑结构的变化。最后给出了仿真结果的理论意义以及模型进一步完善的可能性。

Study the topological structure evolution of neural networks, by assigning an unconnected simulated small neural groups to interact with the vitural environment. In detail, first, the virtual life＇ s small neural group connected with the central neural where no inter-connection, and no connection between central neural and moving neurons which are activated probably, the food are added to the virtual environment. Moving neurons are auto-activited such that relative move produced between life and food,thus lead to Hebb learning and topological structure change of neural network. Finally, the oretical meaning of simulation results and further improvement of the model is discussed.