<p>In this work, we first introduce an alternated and multi-step inertial algorithm (AMiATTCG) for split feasibility problems in real Hilbert spaces. The suggested algorithm employs a three-term conjugate gradient-like direction, enabling faster convergence of its iterates toward a point in the solution set of the problem. The special point of the proposed method in this work is that it is the first method to combine conjugate gradient and general inertial methods in a single algorithm. Additionally, it is the first of its kind in the study of general inertial methods to integrate an alternated inertial term and a multi-step inertial term, both of which are improved versions of the classical inertial term. We establish the strong convergence of AMiATTCG by investigating the convergence of an alternated inertial relaxed algorithm with a three-term conjugate gradient-like direction and perturbations. Notably, our approach avoids some of the restrictive conditions commonly assumed in many methods with conjugate gradient-like directions, and the step-size used does not rely on any information about the norm of the underlying operator or the use of a line search. Moreover, we analyze the applications of the suggested method to classification problems based on the <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="40314_2025_3276_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\ell _{1}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>ℓ</mi> <mn>1</mn> </msub> </math></EquationSource> </InlineEquation>-regularization approach (i.e., the Lasso model) and the <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="40314_2025_3276_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="49" /> </InlineMediaObject> <EquationSource Format="TEX">\(\ell _{1}-\ell _{2}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>ℓ</mi> <mn>1</mn> </msub> <mo>-</mo> <msub> <mi>ℓ</mi> <mn>2</mn> </msub> </mrow> </math></EquationSource> </InlineEquation> hybrid regularization approach. Furthermore, we investigate the efficiency of AMiATTCG in recovering the original signal from a sparse and noisy one using the elastic net model. In all these experiments, the numerical results demonstrate that AMiATTCG is computationally efficient and exhibits greater stability and better generalization performance than some existing algorithms in the literature.</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

Alternated and Multi-Step Inertial Algorithm with Three-Term Conjugate Gradient-Like Direction for Split Feasibilities with Applications in Classification Problems and Elastic Net

  • Abdulwahab Ahmad,
  • Poom Kumam,
  • Mahmoud Muhammad Yahaya,
  • Kanokwan Sitthithakerngkiet

摘要

In this work, we first introduce an alternated and multi-step inertial algorithm (AMiATTCG) for split feasibility problems in real Hilbert spaces. The suggested algorithm employs a three-term conjugate gradient-like direction, enabling faster convergence of its iterates toward a point in the solution set of the problem. The special point of the proposed method in this work is that it is the first method to combine conjugate gradient and general inertial methods in a single algorithm. Additionally, it is the first of its kind in the study of general inertial methods to integrate an alternated inertial term and a multi-step inertial term, both of which are improved versions of the classical inertial term. We establish the strong convergence of AMiATTCG by investigating the convergence of an alternated inertial relaxed algorithm with a three-term conjugate gradient-like direction and perturbations. Notably, our approach avoids some of the restrictive conditions commonly assumed in many methods with conjugate gradient-like directions, and the step-size used does not rely on any information about the norm of the underlying operator or the use of a line search. Moreover, we analyze the applications of the suggested method to classification problems based on the \(\ell _{1}\) 1 -regularization approach (i.e., the Lasso model) and the \(\ell _{1}-\ell _{2}\) 1 - 2 hybrid regularization approach. Furthermore, we investigate the efficiency of AMiATTCG in recovering the original signal from a sparse and noisy one using the elastic net model. In all these experiments, the numerical results demonstrate that AMiATTCG is computationally efficient and exhibits greater stability and better generalization performance than some existing algorithms in the literature.