<p>The goal of this paper is to compare alternative stationarity notions in structured nonsmooth optimization (SNO). Here, nonsmoothness is caused by complementarity, vanishing, orthogonality type, switching, or disjunctive constraints. On one side, we consider geometrically motivated notions of <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10898_2025_1542_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\widehat{N}\)</EquationSource> <EquationSource Format="MATHML"><math> <mover accent="true"> <mi>N</mi> <mo stretchy="true">^</mo> </mover> </math></EquationSource> </InlineEquation>-, <i>N</i>-, and <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10898_2025_1542_Article_IEq2.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\overline{N}\)</EquationSource> <EquationSource Format="MATHML"><math> <mover> <mi>N</mi> <mo>¯</mo> </mover> </math></EquationSource> </InlineEquation>-stationarity in terms of Fréchet, Mordukhovich, and Clarke normal cones to the feasible set, respectively. On the other side, we advocate the notion of topologically relevant T-stationarity, which adequately captures the global structure of SNO. Our main findings say that (a) <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10898_2025_1542_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\widehat{N}\)</EquationSource> <EquationSource Format="MATHML"><math> <mover accent="true"> <mi>N</mi> <mo stretchy="true">^</mo> </mover> </math></EquationSource> </InlineEquation>-stationary points include all local minimizers; (b) <i>N</i>-stationary points, which are not <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10898_2025_1542_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\widehat{N}\)</EquationSource> <EquationSource Format="MATHML"><math> <mover accent="true"> <mi>N</mi> <mo stretchy="true">^</mo> </mover> </math></EquationSource> </InlineEquation>-stationary, correspond to the singular saddle points of first order; (c) T-stationary points, which are not <i>N</i>-stationary, correspond to the regular saddle points of first order; (d) <InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10898_2025_1542_Article_IEq2.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\overline{N}\)</EquationSource> <EquationSource Format="MATHML"><math> <mover> <mi>N</mi> <mo>¯</mo> </mover> </math></EquationSource> </InlineEquation>-stationary points, which are not T-stationary, are irrelevant for optimization purposes, at least from the topological point of view. Overall, a hierarchy of stationarity notions for SNO is established.</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

Stationarity in nonsmooth optimization between geometrical motivation and topological relevance

  • Vladimir Shikhman

摘要

The goal of this paper is to compare alternative stationarity notions in structured nonsmooth optimization (SNO). Here, nonsmoothness is caused by complementarity, vanishing, orthogonality type, switching, or disjunctive constraints. On one side, we consider geometrically motivated notions of \(\widehat{N}\) N ^ -, N-, and \(\overline{N}\) N ¯ -stationarity in terms of Fréchet, Mordukhovich, and Clarke normal cones to the feasible set, respectively. On the other side, we advocate the notion of topologically relevant T-stationarity, which adequately captures the global structure of SNO. Our main findings say that (a) \(\widehat{N}\) N ^ -stationary points include all local minimizers; (b) N-stationary points, which are not \(\widehat{N}\) N ^ -stationary, correspond to the singular saddle points of first order; (c) T-stationary points, which are not N-stationary, correspond to the regular saddle points of first order; (d) \(\overline{N}\) N ¯ -stationary points, which are not T-stationary, are irrelevant for optimization purposes, at least from the topological point of view. Overall, a hierarchy of stationarity notions for SNO is established.