论文部分内容阅读
由文献[2]提出的ART-Ⅱ模型可处理任意序列的辨识问题,但其未涉及网络的动态过程和自稳定区域的研究,只是人工设置不同的ρ值得到仿真结果。本文在深入研究ART-Ⅱ动态反馈机理的基础上,引入了吸引域、自稳定性、聚点等概念,提出了ρ值自适应算法。这一新的机制可以并行搜索和不断地适应外界环境的变化,使得ρ自适应的ART-Ⅱ模型有一定的纠错、容错记忆能力,又有一定的敏感性,克服了原模型ρ值固定时错误记忆的弊病。(由改进型ART-Ⅱ神经网和数据采集、推理机制结合形成一完整智能系统,实现通信系统的多频信号接收处理。系统试验结果表明具有良好的识别效果。
The ART-Ⅱ model proposed by [2] can deal with any sequence identification problem, but it does not involve the dynamic process of the network and the study of self-stable region, but only by manually setting different ρ values to get the simulation results. Based on the deep research on the dynamic feedback mechanism of ART-Ⅱ, the concepts of attracting domain, self-stability and clustering point are introduced and the ρ-adaptive algorithm is proposed. This new mechanism can search in parallel and continuously adapt to the changes of the external environment, making the ρ-adaptive ART-Ⅱ model have some error correction, fault-tolerant memory capability, and a certain sensitivity, which overcomes the fixed ρ value of the original model When the wrong memory of the ills. (The improved ART-Ⅱ neural network and data collection and reasoning mechanism are combined to form a complete intelligent system to realize the multi-frequency signal receiving and processing of the communication system.The system test results show that it has a good recognition effect.