<p>Let <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\{L_k, k\ge 1\}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo stretchy="false">{</mo> <msub> <mi>L</mi> <mi>k</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>≥</mo> <mn>1</mn> <mo stretchy="false">}</mo> </mrow> </math></EquationSource> </InlineEquation> be any increasing unbounded sequence of positive reals and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\((a_k)_{k\ge 1}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mo stretchy="false">(</mo> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo stretchy="false">)</mo> </mrow> <mrow> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </msub> </math></EquationSource> </InlineEquation> a sequence of real numbers such that <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(A_x = \sum _{k\le x} a_k^2\uparrow \infty \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>A</mi> <mi>x</mi> </msub> <mo>=</mo> <msub> <mo>∑</mo> <mrow> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </msub> <msubsup> <mi>a</mi> <mi>k</mi> <mn>2</mn> </msubsup> <mrow> <mo stretchy="false">↑</mo> <mi>∞</mi> </mrow> </mrow> </math></EquationSource> </InlineEquation>, with <i>x</i>. We study moderate deviations of suprema of the Gaussian polynomials <Equation ID="Equ88"> <EquationSource Format="TEX">\(\begin{aligned} X_{y,x}(u) = \sum _{y\le k\le x} a_k \big ( g_k \cos L_ku+ g'_k\sin L_k u\big ),{\qquad }x&gt;y\ge 1,\quad u\in \mathbb {R}, \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <msub> <mi>X</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mrow> <mo stretchy="false">(</mo> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <munder> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </munder> <msub> <mi>a</mi> <mi>k</mi> </msub> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">(</mo> </mrow> <msub> <mi>g</mi> <mi>k</mi> </msub> <mo>cos</mo> <msub> <mi>L</mi> <mi>k</mi> </msub> <mi>u</mi> <mo>+</mo> <msubsup> <mi>g</mi> <mi>k</mi> <mo>′</mo> </msubsup> <mo>sin</mo> <msub> <mi>L</mi> <mi>k</mi> </msub> <mi>u</mi> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">)</mo> </mrow> <mo>,</mo> <mspace width="2em" /> <mi>x</mi> <mo>&gt;</mo> <mi>y</mi> <mo>≥</mo> <mn>1</mn> <mo>,</mo> <mspace width="1em" /> <mi>u</mi> <mo>∈</mo> <mi mathvariant="double-struck">R</mi> <mo>,</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>where <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\((g_k)_{k\ge 1}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mo stretchy="false">(</mo> <msub> <mi>g</mi> <mi>k</mi> </msub> <mo stretchy="false">)</mo> </mrow> <mrow> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </msub> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\((g'_k)_{k\ge 1}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mo stretchy="false">(</mo> <msubsup> <mi>g</mi> <mi>k</mi> <mo>′</mo> </msubsup> <mo stretchy="false">)</mo> </mrow> <mrow> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </msub> </math></EquationSource> </InlineEquation> are two independent sequences of i.i.d. <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(\mathcal N(0,1)\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="script">N</mi> <mo stretchy="false">(</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo stretchy="false">)</mo> </mrow> </math></EquationSource> </InlineEquation> distributed random variables. We first study the periodic case <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(L_k\equiv k\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>L</mi> <mi>k</mi> </msub> <mo>≡</mo> <mi>k</mi> </mrow> </math></EquationSource> </InlineEquation>. Assume for instance that <InlineEquation ID="IEq8"> <EquationSource Format="TEX">\(A(x)\sim \log \log x\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>A</mi> <mo stretchy="false">(</mo> <mi>x</mi> <mo stretchy="false">)</mo> <mo>∼</mo> <mo>log</mo> <mo>log</mo> <mi>x</mi> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq9"> <EquationSource Format="TEX">\(x\rightarrow \infty \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>x</mi> <mo stretchy="false">→</mo> <mi>∞</mi> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq10"> <EquationSource Format="TEX">\(B=\sum _{k\ge 1} a_k^4&lt;\infty \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>B</mi> <mo>=</mo> <msub> <mo>∑</mo> <mrow> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </msub> <msubsup> <mi>a</mi> <mi>k</mi> <mn>4</mn> </msubsup> <mo>&lt;</mo> <mi>∞</mi> </mrow> </math></EquationSource> </InlineEquation>. Let <InlineEquation ID="IEq11"> <EquationSource Format="TEX">\(0&lt;\eta &lt; 1\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0</mn> <mo>&lt;</mo> <mi>η</mi> <mo>&lt;</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation>. We prove that there exists an absolute constant <i>C</i> such that for all <i>x</i> large enough, <Equation ID="Equ89"> <EquationSource Format="TEX">\(\begin{aligned} \mathbb {P}\Big \{ \sup _{0\le t\le 1} X_{1,x}(t) \le \sqrt{2\eta (\log \log x)(\log \log \log x)}\Big \} \ \le \ e^{-\,\frac{C (\log \log x)^{1-\eta }}{ \sqrt{8\eta (B+1)(\log \log \log x)}}}. \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <mi mathvariant="double-struck">P</mi> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">{</mo> </mrow> <munder> <mo movablelimits="true">sup</mo> <mrow> <mn>0</mn> <mo>≤</mo> <mi>t</mi> <mo>≤</mo> <mn>1</mn> </mrow> </munder> <msub> <mi>X</mi> <mrow> <mn>1</mn> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mrow> <mo stretchy="false">(</mo> <mi>t</mi> <mo stretchy="false">)</mo> </mrow> <mo>≤</mo> <msqrt> <mrow> <mn>2</mn> <mi>η</mi> <mo stretchy="false">(</mo> <mo>log</mo> <mo>log</mo> <mi>x</mi> <mo stretchy="false">)</mo> <mo stretchy="false">(</mo> <mo>log</mo> <mo>log</mo> <mo>log</mo> <mi>x</mi> <mo stretchy="false">)</mo> </mrow> </msqrt> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">}</mo> </mrow> <mspace width="4pt" /> <mo>≤</mo> <mspace width="4pt" /> <msup> <mi>e</mi> <mrow> <mo>-</mo> <mspace width="0.166667em" /> <mfrac> <mrow> <mi>C</mi> <msup> <mrow> <mo stretchy="false">(</mo> <mo>log</mo> <mo>log</mo> <mi>x</mi> <mo stretchy="false">)</mo> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <mi>η</mi> </mrow> </msup> </mrow> <msqrt> <mrow> <mn>8</mn> <mi>η</mi> <mo stretchy="false">(</mo> <mi>B</mi> <mo>+</mo> <mn>1</mn> <mo stretchy="false">)</mo> <mo stretchy="false">(</mo> <mo>log</mo> <mo>log</mo> <mo>log</mo> <mi>x</mi> <mo stretchy="false">)</mo> </mrow> </msqrt> </mfrac> </mrow> </msup> <mo>.</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>In the almost periodic case, we prove an approximation theorem. We introduce a modulable diophantine approximation. Let <InlineEquation ID="IEq12"> <EquationSource Format="TEX">\(N_k\ge k, \ k\ge 1\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>N</mi> <mi>k</mi> </msub> <mo>≥</mo> <mi>k</mi> <mo>,</mo> <mspace width="4pt" /> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation> be a non-decreasing unbounded test sequence of positive integers, and let <InlineEquation ID="IEq13"> <EquationSource Format="TEX">\(\ell (k)=\ell (N_k,k)=\frac{1}{N_k}\big \lfloor N_kL_k\big \rfloor \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>ℓ</mi> <mrow> <mo stretchy="false">(</mo> <mi>k</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <mi>ℓ</mi> <mrow> <mo stretchy="false">(</mo> <msub> <mi>N</mi> <mi>k</mi> </msub> <mo>,</mo> <mi>k</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <msub> <mi>N</mi> <mi>k</mi> </msub> </mfrac> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">⌊</mo> </mrow> <msub> <mi>N</mi> <mi>k</mi> </msub> <msub> <mi>L</mi> <mi>k</mi> </msub> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">⌋</mo> </mrow> </mrow> </math></EquationSource> </InlineEquation>, so that <InlineEquation ID="IEq14"> <EquationSource Format="TEX">\(\big |\ell ( k) -L_k\big |\le \frac{1}{N_k}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">|</mo> </mrow> <mi>ℓ</mi> <mrow> <mo stretchy="false">(</mo> <mi>k</mi> <mo stretchy="false">)</mo> </mrow> <mo>-</mo> <msub> <mi>L</mi> <mi>k</mi> </msub> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">|</mo> </mrow> <mo>≤</mo> <mfrac> <mn>1</mn> <msub> <mi>N</mi> <mi>k</mi> </msub> </mfrac> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq15"> <EquationSource Format="TEX">\(k\ge 1\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>k</mi> <mo>≥</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation>. Put for any interval <i>I</i>, <InlineEquation ID="IEq16"> <EquationSource Format="TEX">\( {\kappa }(I)= \#\{{\kappa }: [N_{{\kappa }-1}, N_{\kappa }[\subset I\}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>κ</mi> <mrow> <mo stretchy="false">(</mo> <mi>I</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <mrow> <mo>#</mo> <mo stretchy="false">{</mo> <mi>κ</mi> <mo>:</mo> <mo stretchy="false">[</mo> </mrow> <msub> <mi>N</mi> <mrow> <mi>κ</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>,</mo> <msub> <mi>N</mi> <mi>κ</mi> </msub> <mrow> <mo stretchy="false">[</mo> <mo>⊂</mo> <mi>I</mi> <mo stretchy="false">}</mo> </mrow> </mrow> </math></EquationSource> </InlineEquation>. We prove an approximation theorem by Gaussian polynomials with <InlineEquation ID="IEq17"> <EquationSource Format="TEX">\(\mathbb {Q}\)</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="double-struck">Q</mi> </math></EquationSource> </InlineEquation>-frequencies, and exponentially decaying error term: for any reals <InlineEquation ID="IEq18"> <EquationSource Format="TEX">\(\Theta _{y,x}&gt;0\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi mathvariant="normal">Θ</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mo>&gt;</mo> <mn>0</mn> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq19"> <EquationSource Format="TEX">\(1\le y\le x\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>≤</mo> <mi>y</mi> <mo>≤</mo> <mi>x</mi> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq20"> <EquationSource Format="TEX">\(U\ge 1\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>U</mi> <mo>≥</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq21"> <EquationSource Format="TEX">\(0&lt;h&lt; \Theta _{y,x}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0</mn> <mo>&lt;</mo> <mi>h</mi> <mo>&lt;</mo> <msub> <mi mathvariant="normal">Θ</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> </mrow> </math></EquationSource> </InlineEquation>, <Equation ID="Equ90"> <EquationSource Format="TEX">\(\begin{aligned} \mathbb {P}\Big \{ \sup _{1\le u\le U} X_{y,x}(u)\le \Theta _{y,x} -h \Big \} \, \le \, \mathbb {P}\Big \{ \sup _{1\le u\le U} X^\perp _{y,x}(u) \le \Theta _{y,x} \Big \} +2\, \exp \Big \{\frac{- C\,h^2 }{ {\Delta }^2 \log {\kappa }([1,U]) } \Big \}, \end{aligned}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mtable> <mtr> <mtd columnalign="right"> <mrow> <mi mathvariant="double-struck">P</mi> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">{</mo> </mrow> <munder> <mo movablelimits="true">sup</mo> <mrow> <mn>1</mn> <mo>≤</mo> <mi>u</mi> <mo>≤</mo> <mi>U</mi> </mrow> </munder> <msub> <mi>X</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mrow> <mo stretchy="false">(</mo> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mo>≤</mo> <msub> <mi mathvariant="normal">Θ</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mo>-</mo> <mi>h</mi> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">}</mo> </mrow> <mspace width="0.166667em" /> <mo>≤</mo> <mspace width="0.166667em" /> <mi mathvariant="double-struck">P</mi> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">{</mo> </mrow> <munder> <mo movablelimits="true">sup</mo> <mrow> <mn>1</mn> <mo>≤</mo> <mi>u</mi> <mo>≤</mo> <mi>U</mi> </mrow> </munder> <msubsup> <mi>X</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> <mo>⊥</mo> </msubsup> <mrow> <mo stretchy="false">(</mo> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mo>≤</mo> <msub> <mi mathvariant="normal">Θ</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> </msub> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">}</mo> </mrow> <mo>+</mo> <mn>2</mn> <mspace width="0.166667em" /> <mo>exp</mo> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">{</mo> </mrow> <mfrac> <mrow> <mo>-</mo> <mi>C</mi> <mspace width="0.166667em" /> <msup> <mi>h</mi> <mn>2</mn> </msup> </mrow> <mrow> <msup> <mrow> <mi mathvariant="normal">Δ</mi> </mrow> <mn>2</mn> </msup> <mo>log</mo> <mi>κ</mi> <mrow> <mo stretchy="false">(</mo> <mrow> <mo stretchy="false">[</mo> <mn>1</mn> <mo>,</mo> <mi>U</mi> <mo stretchy="false">]</mo> </mrow> <mo stretchy="false">)</mo> </mrow> </mrow> </mfrac> <mrow> <mo maxsize="1.623em" minsize="1.623em" stretchy="true">}</mo> </mrow> <mo>,</mo> </mrow> </mtd> </mtr> </mtable> </mrow> </math></EquationSource> </Equation>where <InlineEquation ID="IEq22"> <EquationSource Format="TEX">\(X^\perp _{ y,x}(u) = \sum _{ y\le k\le x} a_k \big ( g_k \cos ( \ell ( k) u) \big ) + g'_k\sin ( \ell ( k) u) \big )\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msubsup> <mi>X</mi> <mrow> <mi>y</mi> <mo>,</mo> <mi>x</mi> </mrow> <mo>⊥</mo> </msubsup> <mrow> <mo stretchy="false">(</mo> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mo>=</mo> <msub> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </msub> <msub> <mi>a</mi> <mi>k</mi> </msub> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">(</mo> </mrow> <msub> <mi>g</mi> <mi>k</mi> </msub> <mo>cos</mo> <mrow> <mo stretchy="false">(</mo> <mi>ℓ</mi> <mrow> <mo stretchy="false">(</mo> <mi>k</mi> <mo stretchy="false">)</mo> </mrow> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">)</mo> </mrow> <mo>+</mo> <msubsup> <mi>g</mi> <mi>k</mi> <mo>′</mo> </msubsup> <mo>sin</mo> <mrow> <mo stretchy="false">(</mo> <mi>ℓ</mi> <mrow> <mo stretchy="false">(</mo> <mi>k</mi> <mo stretchy="false">)</mo> </mrow> <mi>u</mi> <mo stretchy="false">)</mo> </mrow> <mrow> <mo maxsize="1.2em" minsize="1.2em" stretchy="true">)</mo> </mrow> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq23"> <EquationSource Format="TEX">\(u\in \mathbb {R}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>u</mi> <mo>∈</mo> <mi mathvariant="double-struck">R</mi> </mrow> </math></EquationSource> </InlineEquation>, and <Equation ID="Equ91"> <EquationSource Format="TEX">\({\Delta }\,=\, \sqrt{ \sum _{ y\le k\le x} \frac{1 }{N^2_k } } \sqrt{ \sum _{ y\le k\le x}a_k^2 } +\underset{\underset{N_{\kappa }\le U}{y\le k&lt; {\kappa }}}{\sum } |a_k| + \sup _{ y\le N_{\kappa }\le U }\ N_{\kappa }\sqrt{ \sum _{ {\kappa }\le k\le x} \frac{ 1}{N_k^{2 } } }\, \sqrt{ \sum _{ {\kappa }\le k\le x} |a_k|^2 },\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mrow> <mi mathvariant="normal">Δ</mi> <mspace width="0.166667em" /> <mo>=</mo> <mspace width="0.166667em" /> <msqrt> <mrow> <munder> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </munder> <mfrac> <mn>1</mn> <msubsup> <mi>N</mi> <mi>k</mi> <mn>2</mn> </msubsup> </mfrac> </mrow> </msqrt> <msqrt> <mrow> <munder> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </munder> <msubsup> <mi>a</mi> <mi>k</mi> <mn>2</mn> </msubsup> </mrow> </msqrt> <mo>+</mo> <munder> <mo>∑</mo> <munder> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>&lt;</mo> <mi>κ</mi> </mrow> <mrow> <msub> <mi>N</mi> <mi>κ</mi> </msub> <mo>≤</mo> <mi>U</mi> </mrow> </munder> </munder> <mrow> <mo stretchy="false">|</mo> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo stretchy="false">|</mo> </mrow> <mo>+</mo> <munder> <mo movablelimits="true">sup</mo> <mrow> <mi>y</mi> <mo>≤</mo> <msub> <mi>N</mi> <mi>κ</mi> </msub> <mo>≤</mo> <mi>U</mi> </mrow> </munder> <mspace width="4pt" /> <msub> <mi>N</mi> <mi>κ</mi> </msub> <msqrt> <mrow> <munder> <mo>∑</mo> <mrow> <mi>κ</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </munder> <mfrac> <mn>1</mn> <msubsup> <mi>N</mi> <mi>k</mi> <mn>2</mn> </msubsup> </mfrac> </mrow> </msqrt> <mspace width="0.166667em" /> <msqrt> <mrow> <munder> <mo>∑</mo> <mrow> <mi>κ</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </munder> <msup> <mrow> <mo stretchy="false">|</mo> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo stretchy="false">|</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> <mo>,</mo> </mrow> </math></EquationSource> </Equation>if <InlineEquation ID="IEq24"> <EquationSource Format="TEX">\(1\le y\le U\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>≤</mo> <mi>y</mi> <mo>≤</mo> <mi>U</mi> </mrow> </math></EquationSource> </InlineEquation>, and <InlineEquation ID="IEq25"> <EquationSource Format="TEX">\({\Delta }\,=\,U \sqrt{ \sum _{ y\le k\le x}\frac{1 }{N^2_k } }\sqrt{ \sum _{ y\le k\le x}a_k^2\ }\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="normal">Δ</mi> <mspace width="0.166667em" /> <mo>=</mo> <mspace width="0.166667em" /> <mi>U</mi> <msqrt> <mrow> <msub> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </msub> <mfrac> <mn>1</mn> <msubsup> <mi>N</mi> <mi>k</mi> <mn>2</mn> </msubsup> </mfrac> </mrow> </msqrt> <msqrt> <mrow> <msub> <mo>∑</mo> <mrow> <mi>y</mi> <mo>≤</mo> <mi>k</mi> <mo>≤</mo> <mi>x</mi> </mrow> </msub> <msubsup> <mi>a</mi> <mi>k</mi> <mn>2</mn> </msubsup> <mspace width="4pt" /> </mrow> </msqrt> </mrow> </math></EquationSource> </InlineEquation>, if <InlineEquation ID="IEq26"> <EquationSource Format="TEX">\(1\le U\le y\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>≤</mo> <mi>U</mi> <mo>≤</mo> <mi>y</mi> </mrow> </math></EquationSource> </InlineEquation>. Finally we study for general non-vanishing coefficient sequences, the behavior along lattices of almost periodic Gaussian polynomials with linearly independent frequencies, and use a lattice localized version of Kronecker’s theorem.</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

Moderate Deviations of Suprema of Gaussian Processes–a Cyclic Approximation Criterion

  • Michel J. G. Weber

摘要

Let \(\{L_k, k\ge 1\}\) { L k , k 1 } be any increasing unbounded sequence of positive reals and \((a_k)_{k\ge 1}\) ( a k ) k 1 a sequence of real numbers such that \(A_x = \sum _{k\le x} a_k^2\uparrow \infty \) A x = k x a k 2 , with x. We study moderate deviations of suprema of the Gaussian polynomials \(\begin{aligned} X_{y,x}(u) = \sum _{y\le k\le x} a_k \big ( g_k \cos L_ku+ g'_k\sin L_k u\big ),{\qquad }x>y\ge 1,\quad u\in \mathbb {R}, \end{aligned}\) X y , x ( u ) = y k x a k ( g k cos L k u + g k sin L k u ) , x > y 1 , u R , where \((g_k)_{k\ge 1}\) ( g k ) k 1 , \((g'_k)_{k\ge 1}\) ( g k ) k 1 are two independent sequences of i.i.d. \(\mathcal N(0,1)\) N ( 0 , 1 ) distributed random variables. We first study the periodic case \(L_k\equiv k\) L k k . Assume for instance that \(A(x)\sim \log \log x\) A ( x ) log log x , \(x\rightarrow \infty \) x and \(B=\sum _{k\ge 1} a_k^4<\infty \) B = k 1 a k 4 < . Let \(0<\eta < 1\) 0 < η < 1 . We prove that there exists an absolute constant C such that for all x large enough, \(\begin{aligned} \mathbb {P}\Big \{ \sup _{0\le t\le 1} X_{1,x}(t) \le \sqrt{2\eta (\log \log x)(\log \log \log x)}\Big \} \ \le \ e^{-\,\frac{C (\log \log x)^{1-\eta }}{ \sqrt{8\eta (B+1)(\log \log \log x)}}}. \end{aligned}\) P { sup 0 t 1 X 1 , x ( t ) 2 η ( log log x ) ( log log log x ) } e - C ( log log x ) 1 - η 8 η ( B + 1 ) ( log log log x ) . In the almost periodic case, we prove an approximation theorem. We introduce a modulable diophantine approximation. Let \(N_k\ge k, \ k\ge 1\) N k k , k 1 be a non-decreasing unbounded test sequence of positive integers, and let \(\ell (k)=\ell (N_k,k)=\frac{1}{N_k}\big \lfloor N_kL_k\big \rfloor \) ( k ) = ( N k , k ) = 1 N k N k L k , so that \(\big |\ell ( k) -L_k\big |\le \frac{1}{N_k}\) | ( k ) - L k | 1 N k , \(k\ge 1\) k 1 . Put for any interval I, \( {\kappa }(I)= \#\{{\kappa }: [N_{{\kappa }-1}, N_{\kappa }[\subset I\}\) κ ( I ) = # { κ : [ N κ - 1 , N κ [ I } . We prove an approximation theorem by Gaussian polynomials with \(\mathbb {Q}\) Q -frequencies, and exponentially decaying error term: for any reals \(\Theta _{y,x}>0\) Θ y , x > 0 , \(1\le y\le x\) 1 y x , \(U\ge 1\) U 1 , \(0<h< \Theta _{y,x}\) 0 < h < Θ y , x , \(\begin{aligned} \mathbb {P}\Big \{ \sup _{1\le u\le U} X_{y,x}(u)\le \Theta _{y,x} -h \Big \} \, \le \, \mathbb {P}\Big \{ \sup _{1\le u\le U} X^\perp _{y,x}(u) \le \Theta _{y,x} \Big \} +2\, \exp \Big \{\frac{- C\,h^2 }{ {\Delta }^2 \log {\kappa }([1,U]) } \Big \}, \end{aligned}\) P { sup 1 u U X y , x ( u ) Θ y , x - h } P { sup 1 u U X y , x ( u ) Θ y , x } + 2 exp { - C h 2 Δ 2 log κ ( [ 1 , U ] ) } , where \(X^\perp _{ y,x}(u) = \sum _{ y\le k\le x} a_k \big ( g_k \cos ( \ell ( k) u) \big ) + g'_k\sin ( \ell ( k) u) \big )\) X y , x ( u ) = y k x a k ( g k cos ( ( k ) u ) ) + g k sin ( ( k ) u ) ) , \(u\in \mathbb {R}\) u R , and \({\Delta }\,=\, \sqrt{ \sum _{ y\le k\le x} \frac{1 }{N^2_k } } \sqrt{ \sum _{ y\le k\le x}a_k^2 } +\underset{\underset{N_{\kappa }\le U}{y\le k< {\kappa }}}{\sum } |a_k| + \sup _{ y\le N_{\kappa }\le U }\ N_{\kappa }\sqrt{ \sum _{ {\kappa }\le k\le x} \frac{ 1}{N_k^{2 } } }\, \sqrt{ \sum _{ {\kappa }\le k\le x} |a_k|^2 },\) Δ = y k x 1 N k 2 y k x a k 2 + y k < κ N κ U | a k | + sup y N κ U N κ κ k x 1 N k 2 κ k x | a k | 2 , if \(1\le y\le U\) 1 y U , and \({\Delta }\,=\,U \sqrt{ \sum _{ y\le k\le x}\frac{1 }{N^2_k } }\sqrt{ \sum _{ y\le k\le x}a_k^2\ }\) Δ = U y k x 1 N k 2 y k x a k 2 , if \(1\le U\le y\) 1 U y . Finally we study for general non-vanishing coefficient sequences, the behavior along lattices of almost periodic Gaussian polynomials with linearly independent frequencies, and use a lattice localized version of Kronecker’s theorem.