Scholarlink

<p>This article introduces a two-stage training strategy for spiking neural networks, which integrates an adaptive decay time algorithm with a normalized spiking error backpropagation algorithm to reduce the network output error more quickly and stably. This training strategy is applied to identify the aerodynamic derivatives of an aircraft. The reliability of these aerodynamic derivatives is assessed using the bootstrapping technique, which serves as a basis for verifying the effectiveness of the proposed approach. Simulation results demonstrate that the output error reduction speed and convergence speed of the proposed method outperforms traditional methods. Additionally, parameters identified using this strategy are more reliable than those obtained from other popular approaches.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Proposing a two-stage training strategy for spike neural networks to enhance the efficiency of aircraft aerodynamic derivative identification

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Nguyen Van Tuan,
  • Truong Dang Khoa,
  • Pham Trung Dung

摘要

This article introduces a two-stage training strategy for spiking neural networks, which integrates an adaptive decay time algorithm with a normalized spiking error backpropagation algorithm to reduce the network output error more quickly and stably. This training strategy is applied to identify the aerodynamic derivatives of an aircraft. The reliability of these aerodynamic derivatives is assessed using the bootstrapping technique, which serves as a basis for verifying the effectiveness of the proposed approach. Simulation results demonstrate that the output error reduction speed and convergence speed of the proposed method outperforms traditional methods. Additionally, parameters identified using this strategy are more reliable than those obtained from other popular approaches.