<p>This paper considers the optimization problem <Equation ID="Equ63"> <MediaObject> <ImageObject Color="BlackWhite" FileRef="10915_2025_2806_Article_Equ63.gif" Format="GIF" Height="28" Rendition="HTML" Resolution="72" Type="Linedraw" Width="149" /> </MediaObject> <EquationSource Format="TEX">\(\begin{aligned} \min _{X \in \mathcal {F}_v} f( {X}) + \lambda \Vert X\Vert _1, \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <munder> <mo movablelimits="true">min</mo> <mrow> <mi>X</mi> <mo>∈</mo> <msub> <mi mathvariant="script">F</mi> <mi>v</mi> </msub> </mrow> </munder> <mi>f</mi> <mrow> <mo stretchy="false">(</mo> <mi>X</mi> <mo stretchy="false">)</mo> </mrow> <mo>+</mo> <mi>λ</mi> <msub> <mrow> <mo stretchy="false">‖</mo> <mi>X</mi> <mo stretchy="false">‖</mo> </mrow> <mn>1</mn> </msub> <mo>,</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>where <i>f</i> is smooth, <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10915_2025_2806_Article_IEq1.gif" Format="GIF" Height="21" Rendition="HTML" Resolution="72" Type="Linedraw" Width="319" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mathcal {F}_v = \{X \in \mathbb {R}^{n \times q}: X^T X = I_q, v \in \textrm{span}(X)\}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi mathvariant="script">F</mi> <mi>v</mi> </msub> <mo>=</mo> <mrow> <mo stretchy="false">{</mo> <mi>X</mi> <mo>∈</mo> <msup> <mrow> <mi mathvariant="double-struck">R</mi> </mrow> <mrow> <mi>n</mi> <mo>×</mo> <mi>q</mi> </mrow> </msup> <mo>:</mo> <msup> <mi>X</mi> <mi>T</mi> </msup> <mi>X</mi> <mo>=</mo> <msub> <mi>I</mi> <mi>q</mi> </msub> <mo>,</mo> <mi>v</mi> <mo>∈</mo> <mtext>span</mtext> <mrow> <mo stretchy="false">(</mo> <mi>X</mi> <mo stretchy="false">)</mo> </mrow> <mo stretchy="false">}</mo> </mrow> </mrow> </math></EquationSource> </InlineEquation>, and <i>v</i> is a given positive vector. The clustering models including but not limited to the models used by <i>k</i>-means, community detection, and normalized cut can be reformulated as such optimization problems. It is proven that the domain <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10915_2025_2806_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mathcal {F}_v\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi mathvariant="script">F</mi> <mi>v</mi> </msub> </math></EquationSource> </InlineEquation> forms a compact embedded submanifold of <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10915_2025_2806_Article_IEq3.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="35" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mathbb {R}^{n \times q}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi mathvariant="double-struck">R</mi> </mrow> <mrow> <mi>n</mi> <mo>×</mo> <mi>q</mi> </mrow> </msup> </math></EquationSource> </InlineEquation> and optimization-related tools including a family of computationally efficient retractions and an orthonormal basis of any normal space of <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10915_2025_2806_Article_IEq4.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mathcal {F}_v\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi mathvariant="script">F</mi> <mi>v</mi> </msub> </math></EquationSource> </InlineEquation> are derived. A Riemannian proximal gradient method that allows an adaptive step size is proposed. The proposed Riemannian proximal gradient method solves its subproblem inexactly and still guarantees its global convergence. Numerical experiments on community detection in networks and normalized cut for image segmentation are used to demonstrate the performance of the proposed method.</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

A Riemannian Optimization Approach to Clustering Problems

  • Wen Huang,
  • Meng Wei,
  • Kyle A. Gallivan,
  • Paul Van Dooren

摘要

This paper considers the optimization problem \(\begin{aligned} \min _{X \in \mathcal {F}_v} f( {X}) + \lambda \Vert X\Vert _1, \end{aligned}\) min X F v f ( X ) + λ X 1 , where f is smooth, \(\mathcal {F}_v = \{X \in \mathbb {R}^{n \times q}: X^T X = I_q, v \in \textrm{span}(X)\}\) F v = { X R n × q : X T X = I q , v span ( X ) } , and v is a given positive vector. The clustering models including but not limited to the models used by k-means, community detection, and normalized cut can be reformulated as such optimization problems. It is proven that the domain \(\mathcal {F}_v\) F v forms a compact embedded submanifold of \(\mathbb {R}^{n \times q}\) R n × q and optimization-related tools including a family of computationally efficient retractions and an orthonormal basis of any normal space of \(\mathcal {F}_v\) F v are derived. A Riemannian proximal gradient method that allows an adaptive step size is proposed. The proposed Riemannian proximal gradient method solves its subproblem inexactly and still guarantees its global convergence. Numerical experiments on community detection in networks and normalized cut for image segmentation are used to demonstrate the performance of the proposed method.