<p>We construct an analogue of Dyson Brownian motion in the Siegel half-space <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\mathcal {H}\)</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="script">H</mi> </math></EquationSource> </InlineEquation> that we term <i>Siegel Brownian motion</i>. Given <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\beta \in (0,\infty ]\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>β</mi> <mo>∈</mo> <mo stretchy="false">(</mo> <mn>0</mn> <mo>,</mo> <mi>∞</mi> <mo stretchy="false">]</mo> </mrow> </math></EquationSource> </InlineEquation>, a stochastic flow for <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(Z_t\in \mathcal {H}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>Z</mi> <mi>t</mi> </msub> <mo>∈</mo> <mi mathvariant="script">H</mi> </mrow> </math></EquationSource> </InlineEquation> is introduced so that the law of the eigenvalues <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\lambda _t\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>λ</mi> <mi>t</mi> </msub> </math></EquationSource> </InlineEquation> of the cross ratio matrix <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\({\mathfrak {R}}(Z_t,\varvec{i}I_n)\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="fraktur">R</mi> <mo stretchy="false">(</mo> <msub> <mi>Z</mi> <mi>t</mi> </msub> <mo>,</mo> <mrow> <mi mathvariant="bold-italic">i</mi> </mrow> <msub> <mi>I</mi> <mi>n</mi> </msub> <mo stretchy="false">)</mo> </mrow> </math></EquationSource> </InlineEquation> is determined, after a change of variables to <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(\sigma (\lambda ) \in (0,\infty )^n\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>σ</mi> <mrow> <mo stretchy="false">(</mo> <mi>λ</mi> <mo stretchy="false">)</mo> </mrow> <mo>∈</mo> <msup> <mrow> <mo stretchy="false">(</mo> <mn>0</mn> <mo>,</mo> <mi>∞</mi> <mo stretchy="false">)</mo> </mrow> <mi>n</mi> </msup> </mrow> </math></EquationSource> </InlineEquation>, by the Itô differential equation <Equation ID="Equ1"> <EquationNumber>0.1</EquationNumber> <EquationSource Format="TEX">\(\begin{aligned} \textrm{d}\sigma ^k_t=\frac{1}{2}\left( \coth {\sigma ^k_t}+\sum _{l\ne k}\frac{\sinh {\sigma ^k_t}}{\cosh {\sigma ^k_t}-\cosh {\sigma ^l_t}}\right) \textrm{d}t+\sqrt{\frac{2}{\beta }}\textrm{d}W^k_t, \quad k=1,\ldots , n, \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <mtext>d</mtext> <msubsup> <mi>σ</mi> <mi>t</mi> <mi>k</mi> </msubsup> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mfenced close=")" open="("> <mo>coth</mo> <msubsup> <mi>σ</mi> <mi>t</mi> <mi>k</mi> </msubsup> <mo>+</mo> <munder> <mo>∑</mo> <mrow> <mi>l</mi> <mo>≠</mo> <mi>k</mi> </mrow> </munder> <mfrac> <mrow> <mo>sinh</mo> <msubsup> <mi>σ</mi> <mi>t</mi> <mi>k</mi> </msubsup> </mrow> <mrow> <mo>cosh</mo> <msubsup> <mi>σ</mi> <mi>t</mi> <mi>k</mi> </msubsup> <mo>-</mo> <mo>cosh</mo> <msubsup> <mi>σ</mi> <mi>t</mi> <mi>l</mi> </msubsup> </mrow> </mfrac> </mfenced> <mtext>d</mtext> <mi>t</mi> <mo>+</mo> <msqrt> <mfrac> <mn>2</mn> <mi>β</mi> </mfrac> </msqrt> <mtext>d</mtext> <msubsup> <mi>W</mi> <mi>t</mi> <mi>k</mi> </msubsup> <mo>,</mo> <mspace width="1em" /> <mi>k</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mi>n</mi> <mo>,</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>where <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(W_t\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>W</mi> <mi>t</mi> </msub> </math></EquationSource> </InlineEquation> is a standard Wiener process in <InlineEquation ID="IEq8"> <EquationSource Format="TEX">\(\mathbb {R}^n\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi mathvariant="double-struck">R</mi> </mrow> <mi>n</mi> </msup> </math></EquationSource> </InlineEquation>. This interacting particle system corresponds to stochastic gradient ascent <Equation ID="Equ2"> <EquationNumber>0.2</EquationNumber> <EquationSource Format="TEX">\(\begin{aligned} \textrm{d}\sigma _t= \frac{1}{2}\nabla S(\sigma _t) +\sqrt{\frac{2}{\beta }}\textrm{d}W_t, \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <mtext>d</mtext> <msub> <mi>σ</mi> <mi>t</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mi mathvariant="normal">∇</mi> <mi>S</mi> <mrow> <mo stretchy="false">(</mo> <msub> <mi>σ</mi> <mi>t</mi> </msub> <mo stretchy="false">)</mo> </mrow> <mo>+</mo> <msqrt> <mfrac> <mn>2</mn> <mi>β</mi> </mfrac> </msqrt> <mtext>d</mtext> <msub> <mi>W</mi> <mi>t</mi> </msub> <mo>,</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>where <InlineEquation ID="IEq9"> <EquationSource Format="TEX">\(S(\sigma )= \log \textrm{vol}\,\mathcal {O}_{\lambda (\sigma )}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>S</mi> <mrow> <mo stretchy="false">(</mo> <mi>σ</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <mo>log</mo> <mtext>vol</mtext> <mspace width="0.166667em" /> <msub> <mi mathvariant="script">O</mi> <mrow> <mi>λ</mi> <mo stretchy="false">(</mo> <mi>σ</mi> <mo stretchy="false">)</mo> </mrow> </msub> </mrow> </math></EquationSource> </InlineEquation> is a Boltzmann entropy that enumerates the microstates in the group orbit <InlineEquation ID="IEq10"> <EquationSource Format="TEX">\(\mathcal {O}_\lambda = \{Z \in \mathcal {H}\left| \textrm{eig}\left( {\mathfrak {R}}(Z,\varvec{i}I_n)\right) =\lambda \right. \}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi mathvariant="script">O</mi> <mi>λ</mi> </msub> <mo>=</mo> <mrow> <mo stretchy="false">{</mo> <mi>Z</mi> <mo>∈</mo> <mi mathvariant="script">H</mi> <mfenced open="|"> <mtext>eig</mtext> <mfenced close=")" open="("> <mi mathvariant="fraktur">R</mi> <mo stretchy="false">(</mo> <mi>Z</mi> <mo>,</mo> <mrow> <mi mathvariant="bold-italic">i</mi> </mrow> <msub> <mi>I</mi> <mi>n</mi> </msub> <mo stretchy="false">)</mo> </mfenced> <mo>=</mo> <mi>λ</mi> </mfenced> <mo stretchy="false">}</mo> </mrow> </mrow> </math></EquationSource> </InlineEquation>. In the limit <InlineEquation ID="IEq11"> <EquationSource Format="TEX">\(\beta =\infty \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>β</mi> <mo>=</mo> <mi>∞</mi> </mrow> </math></EquationSource> </InlineEquation>, the group orbits <InlineEquation ID="IEq12"> <EquationSource Format="TEX">\(\mathcal {O}_{\lambda _t}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi mathvariant="script">O</mi> <msub> <mi>λ</mi> <mi>t</mi> </msub> </msub> </math></EquationSource> </InlineEquation> evolve by motion by minus a half times mean curvature.</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

Siegel Brownian motion

  • Govind Menon,
  • Tianmin Yu

摘要

We construct an analogue of Dyson Brownian motion in the Siegel half-space \(\mathcal {H}\) H that we term Siegel Brownian motion. Given \(\beta \in (0,\infty ]\) β ( 0 , ] , a stochastic flow for \(Z_t\in \mathcal {H}\) Z t H is introduced so that the law of the eigenvalues \(\lambda _t\) λ t of the cross ratio matrix \({\mathfrak {R}}(Z_t,\varvec{i}I_n)\) R ( Z t , i I n ) is determined, after a change of variables to \(\sigma (\lambda ) \in (0,\infty )^n\) σ ( λ ) ( 0 , ) n , by the Itô differential equation 0.1 \(\begin{aligned} \textrm{d}\sigma ^k_t=\frac{1}{2}\left( \coth {\sigma ^k_t}+\sum _{l\ne k}\frac{\sinh {\sigma ^k_t}}{\cosh {\sigma ^k_t}-\cosh {\sigma ^l_t}}\right) \textrm{d}t+\sqrt{\frac{2}{\beta }}\textrm{d}W^k_t, \quad k=1,\ldots , n, \end{aligned}\) d σ t k = 1 2 coth σ t k + l k sinh σ t k cosh σ t k - cosh σ t l d t + 2 β d W t k , k = 1 , , n , where \(W_t\) W t is a standard Wiener process in \(\mathbb {R}^n\) R n . This interacting particle system corresponds to stochastic gradient ascent 0.2 \(\begin{aligned} \textrm{d}\sigma _t= \frac{1}{2}\nabla S(\sigma _t) +\sqrt{\frac{2}{\beta }}\textrm{d}W_t, \end{aligned}\) d σ t = 1 2 S ( σ t ) + 2 β d W t , where \(S(\sigma )= \log \textrm{vol}\,\mathcal {O}_{\lambda (\sigma )}\) S ( σ ) = log vol O λ ( σ ) is a Boltzmann entropy that enumerates the microstates in the group orbit \(\mathcal {O}_\lambda = \{Z \in \mathcal {H}\left| \textrm{eig}\left( {\mathfrak {R}}(Z,\varvec{i}I_n)\right) =\lambda \right. \}\) O λ = { Z H eig R ( Z , i I n ) = λ } . In the limit \(\beta =\infty \) β = , the group orbits \(\mathcal {O}_{\lambda _t}\) O λ t evolve by motion by minus a half times mean curvature.