Scholarlink

<p>This work presents a new 4-D hyperchaotic Pan-type system (N4DHPS) characterized by a curve of equilibrium points and examines its associated dynamical behaviors. Hyperchaos is confirmed through numerical analyses, including Lyapunov exponent computation and bifurcation studies. The system is further realized on a Field Programmable Gate Array (FPGA), demonstrating its suitability for high-speed implementation and potential secure communication use. Moreover, a deep learning framework is applied to classify time-series representations of the system’s state variables using transfer learning. A dataset of 4000 images derived from the time-series outputs is evaluated using five pre-trained Deep Learning Models—SqueezeNet, ResNet50, DenseNet201, ShuffleNet, and InceptionResNetV2—where SqueezeNet and ShuffleNet achieve the highest accuracy of 99.34%. The findings highlight the capability of Artificial Intelligence (AI) to reliably identify chaotic patterns, supporting future developments in secure communication and nonlinear dynamical system analysis.</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

Deep learning-based classification and FPGA implementation of a new 4-D hyperchaotic pan system with curve equilibrium point

  • 期刊论文

  • 发表日期：

    2026年4月14日

查看全文
  • Aceng Sambas,
  • Sundarapandian Vaidyanathan,
  • Omar Guillén-Fernández,
  • Esteban Tlelo-Cuautle,
  • Sezgin Kaçar,
  • Süleyman Uzun,
  • Muhammed Telçeken,
  • Khaled Benkouider

摘要

This work presents a new 4-D hyperchaotic Pan-type system (N4DHPS) characterized by a curve of equilibrium points and examines its associated dynamical behaviors. Hyperchaos is confirmed through numerical analyses, including Lyapunov exponent computation and bifurcation studies. The system is further realized on a Field Programmable Gate Array (FPGA), demonstrating its suitability for high-speed implementation and potential secure communication use. Moreover, a deep learning framework is applied to classify time-series representations of the system’s state variables using transfer learning. A dataset of 4000 images derived from the time-series outputs is evaluated using five pre-trained Deep Learning Models—SqueezeNet, ResNet50, DenseNet201, ShuffleNet, and InceptionResNetV2—where SqueezeNet and ShuffleNet achieve the highest accuracy of 99.34%. The findings highlight the capability of Artificial Intelligence (AI) to reliably identify chaotic patterns, supporting future developments in secure communication and nonlinear dynamical system analysis.