Scholarlink

<p>The discrete neuron map models require no discretization process, and they provide favorable design for digital equipment-based neuromorphic systems compared to their continuous counterparts. However, these discrete models include certain nonlinear functions and they have a difficult implementation process with digital equipment. To address this challenge, appropriate function transformation methods are employed to facilitate their hardware realization. This study deals with the Chialvo neuron map model. Its characteristic nonlinear function is simplified to make it suitable for digital hardware implementation. In this process, the modified model must best reflect the dynamic behavior of the original one and the error must be determined as accurately as possible. Based on this requirement, the error between the original and modified CN models is calculated by utilizing their spike timing differences and these differences are determined using the “median&#xa0;+ K ×&#xa0;Median Absolute Deviation (MAD)” method. On the other hand, since the simplification is required to enable easier implementation with digital hardware, it is necessary to demonstrate the feasibility of the implementation and to determine the associated hardware cost. For this purpose, the modified CN model is implemented using a FPGA device. The electrical output signals are recorded with an oscilloscope, and these signals are analyzed through the “Recurrence Quantization Analysis (RQA)” method to assess the neural diversity of the modified model. Ultimately, RQA provides an alternative framework for observing dynamic neural diversity, and this study aims to promote its use in neural and neuromorphic systems.</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

Neural dynamic diversity assessment via recurrence quantification analysis in a hardware-efficient FPGA implementation of the Chialvo neuron map

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Nimet Korkmaz

摘要

The discrete neuron map models require no discretization process, and they provide favorable design for digital equipment-based neuromorphic systems compared to their continuous counterparts. However, these discrete models include certain nonlinear functions and they have a difficult implementation process with digital equipment. To address this challenge, appropriate function transformation methods are employed to facilitate their hardware realization. This study deals with the Chialvo neuron map model. Its characteristic nonlinear function is simplified to make it suitable for digital hardware implementation. In this process, the modified model must best reflect the dynamic behavior of the original one and the error must be determined as accurately as possible. Based on this requirement, the error between the original and modified CN models is calculated by utilizing their spike timing differences and these differences are determined using the “median + K × Median Absolute Deviation (MAD)” method. On the other hand, since the simplification is required to enable easier implementation with digital hardware, it is necessary to demonstrate the feasibility of the implementation and to determine the associated hardware cost. For this purpose, the modified CN model is implemented using a FPGA device. The electrical output signals are recorded with an oscilloscope, and these signals are analyzed through the “Recurrence Quantization Analysis (RQA)” method to assess the neural diversity of the modified model. Ultimately, RQA provides an alternative framework for observing dynamic neural diversity, and this study aims to promote its use in neural and neuromorphic systems.