用于构建人工大脑的神经网络模型的分析与评估  

Analysis and evaluation of neural network model for building artificial brain

作　　者：赵春风[1] 胡政[1] 

机构地区：[1]华中科技大学生命科学与技术学院,武汉430074

出　　处：《计算机应用》2009年第B06期326-328,共3页journal of Computer Applications

摘　　要：目前人工脑的研究还处于起步阶段,构造智能化人工脑的方法正在探索中。影响人工脑性能的关键部分在于所选用的人工神经网络,针对目前已提出的三个网络模型,即CoDi模型、TiPo模型和DePo模型,进行了评估研究。采用的评估方法是通过解决曲线跟踪问题对模型进行测试。测试结果显示DePo模型曲线跟踪取得的效果较另两个更好,TiPo模型跟CoDi模型的性能相似。人工脑的进一步研究工作将包括提出更接近生物机制的模型或工程角度更有进化能力的模型。At present, the artificial brain research is still at its initial period, and the methods to construct artificial brain are being explored. The key part determining the performance of artificial brain is the neural network evolved by genetic algorithm. This paper studied the performance of three models that had been proposed for building artificial brain, namely, CoDi model, TiPo model and DePo model. The models were compared by solving the curve following problems. Experiment results show that DePo model performs better than CoDi model and TiPo model. CoDi model is worst but simple. The further work in building artificial brain would be to develop models that are more consistent with biology mechanism, or to find more evolvable models in engineering respect.

关 键 词：人工脑 神经网络 遗传算法 曲线跟踪 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]