Scholarlink

<p>This paper proposes a deep residual partial least squares regression (PLSR) model, integrated with manifold optimization and Gaussian filter, termed DEGPLSRM. This model addresses the limitations of traditional PLSR. Specifically, traditional PLSR encounters difficulties in handling noisy data effectively and capturing high-level features from complex image datasets. By leveraging the residual network structure and Gaussian filter, DEGPLSRM preserves useful information during feature extraction and smooths image data, effectively reducing the impact of noise. Experimental results on seven distinct datasets demonstrate that DEGPLSRM achieves lower classification error rates and exhibits superior robustness compared to other representative methods. The classification error rates are reduced by up to 71<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2024_3773_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation> on the COIL-20 dataset and by at least 34<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2024_3773_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation> on the Brodatz dataset. The source code and datasets are publicly available to facilitate reproducibility and further research in this domain. The experimental code is available at <a href="https://github.com/wu-xiaobai/degpls.git">https://github.com/wu-xiaobai/degpls.git</a>.</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 residual PLSR model with manifold optimization and Gaussian filter for enhanced image classification

  • 期刊论文

  • 发表日期：

    2025年1月13日

查看全文
  • Xiao Li,
  • Kai Wu,
  • Haoran Chen,
  • Wenjun Song,
  • Hongwei Tao,
  • Zuhe Li,
  • Yanan Du

摘要

This paper proposes a deep residual partial least squares regression (PLSR) model, integrated with manifold optimization and Gaussian filter, termed DEGPLSRM. This model addresses the limitations of traditional PLSR. Specifically, traditional PLSR encounters difficulties in handling noisy data effectively and capturing high-level features from complex image datasets. By leveraging the residual network structure and Gaussian filter, DEGPLSRM preserves useful information during feature extraction and smooths image data, effectively reducing the impact of noise. Experimental results on seven distinct datasets demonstrate that DEGPLSRM achieves lower classification error rates and exhibits superior robustness compared to other representative methods. The classification error rates are reduced by up to 71 \(\%\) % on the COIL-20 dataset and by at least 34 \(\%\) % on the Brodatz dataset. The source code and datasets are publicly available to facilitate reproducibility and further research in this domain. The experimental code is available at https://github.com/wu-xiaobai/degpls.git.