Positive-definite function

<h2 id="definition-1">Definition 1</h2>
Let 
 
 
 
 
 R
 
 
 
 {\displaystyle \mathbb {R} }
 
 be the set of <a href="/facts/Real_number/R02gw5Pb">real numbers</a> and 
 
 
 
 
 C
 
 
 
 {\displaystyle \mathbb {C} }
 
 be the set of <a href="/facts/Complex_number/2w7mqI7e">complex numbers</a>.
A function 
 
 
 
 f
 :
 
 R
 
 →
 
 C
 
 
 
 {\displaystyle f:\mathbb {R} \to \mathbb {C} }
 
 is called positive semi-definite if for all real numbers x1, …, xn the n × n <a href="/facts/Matrix_(mathematics)/qa8FI4ko">matrix</a>

A
        =
        
          
            (
            
              a
              
                i
                j
              
            
            )
          
          
            i
            ,
            j
            =
            1
          
          
            n
          
        
         
        ,
        
        
          a
          
            i
            j
          
        
        =
        f
        (
        
          x
          
            i
          
        
        −
        
          x
          
            j
          
        
        )
      
    
    {\displaystyle A=\left(a_{ij}\right)_{i,j=1}^{n}~,\quad a_{ij}=f(x_{i}-x_{j})}

is a <a href="/facts/Positive-definite_matrix/Hbr6AuPS">positive semi-definite matrix</a>.
By definition, a positive semi-definite matrix, such as 
 
 
 
 A
 
 
 {\displaystyle A}
 
, is <a href="/facts/Hermitian_matrix/qArPblYo">Hermitian</a>; therefore f(−x) is the <a href="/facts/Complex_conjugate/7TEaoQDe">complex conjugate</a> of f(x)).
In particular, it is necessary (but not sufficient) that

f
        (
        0
        )
        ≥
        0
         
        ,
        
        
          |
        
        f
        (
        x
        )
        
          |
        
        ≤
        f
        (
        0
        )
      
    
    {\displaystyle f(0)\geq 0~,\quad |f(x)|\leq f(0)}

(these inequalities follow from the condition for n = 1, 2.)
A function is negative semi-definite if the inequality is reversed. A function is definite if the weak inequality is replaced with a strong (<, > 0).

<h3>Examples</h3>
If 
 
 
 
 (
 X
 ,
 ⟨
 ⋅
 ,
 ⋅
 ⟩
 )
 
 
 {\displaystyle (X,\langle \cdot ,\cdot \rangle )}
 
 is a real <a href="/facts/Inner_product_space/HoyjElEy">inner product space</a>, then 
 
 
 
 
 g
 
 y
 
 
 :
 X
 →
 
 C
 
 
 
 {\displaystyle g_{y}\colon X\to \mathbb {C} }
 
, 
 
 
 
 x
 ↦
 exp
 ⁡
 (
 i
 ⟨
 y
 ,
 x
 ⟩
 )
 
 
 {\displaystyle x\mapsto \exp(i\langle y,x\rangle )}
 
 is positive definite for every 
 
 
 
 y
 ∈
 X
 
 
 {\displaystyle y\in X}
 
: for all 
 
 
 
 u
 ∈
 
 
 C
 
 
 n
 
 
 
 
 {\displaystyle u\in \mathbb {C} ^{n}}
 
 and all 
 
 
 
 
 x
 
 1
 
 
 ,
 …
 ,
 
 x
 
 n
 
 
 
 
 {\displaystyle x_{1},\ldots ,x_{n}}
 
 we have

u
          
            ∗
          
        
        
          A
          
            (
            
              g
              
                y
              
            
            )
          
        
        u
        =
        
          ∑
          
            j
            ,
            k
            =
            1
          
          
            n
          
        
        
          
            
              u
              
                k
              
            
            ¯
          
        
        
          u
          
            j
          
        
        
          e
          
            i
            ⟨
            y
            ,
            
              x
              
                k
              
            
            −
            
              x
              
                j
              
            
            ⟩
          
        
        =
        
          ∑
          
            k
            =
            1
          
          
            n
          
        
        
          
            
              u
              
                k
              
            
            ¯
          
        
        
          e
          
            i
            ⟨
            y
            ,
            
              x
              
                k
              
            
            ⟩
          
        
        
          ∑
          
            j
            =
            1
          
          
            n
          
        
        
          u
          
            j
          
        
        
          e
          
            −
            i
            ⟨
            y
            ,
            
              x
              
                j
              
            
            ⟩
          
        
        =
        
          
            |
            
              
                ∑
                
                  j
                  =
                  1
                
                
                  n
                
              
              
                
                  
                    u
                    
                      j
                    
                  
                  ¯
                
              
              
                e
                
                  i
                  ⟨
                  y
                  ,
                  
                    x
                    
                      j
                    
                  
                  ⟩
                
              
            
            |
          
          
            2
          
        
        ≥
        0.
      
    
    {\displaystyle u^{*}A^{(g_{y})}u=\sum _{j,k=1}^{n}{\overline {u_{k}}}u_{j}e^{i\langle y,x_{k}-x_{j}\rangle }=\sum _{k=1}^{n}{\overline {u_{k}}}e^{i\langle y,x_{k}\rangle }\sum _{j=1}^{n}u_{j}e^{-i\langle y,x_{j}\rangle }=\left|\sum _{j=1}^{n}{\overline {u_{j}}}e^{i\langle y,x_{j}\rangle }\right|^{2}\geq 0.}

As nonnegative linear combinations of positive definite functions are again positive definite, the <a href="/facts/Cosine_function/czej7ur0">cosine function</a> is positive definite as a nonnegative linear combination of the above functions:

cos
        ⁡
        (
        x
        )
        =
        
          
            1
            2
          
        
        (
        
          e
          
            i
            x
          
        
        +
        
          e
          
            −
            i
            x
          
        
        )
        =
        
          
            1
            2
          
        
        (
        
          g
          
            1
          
        
        +
        
          g
          
            −
            1
          
        
        )
        .
      
    
    {\displaystyle \cos(x)={\frac {1}{2}}(e^{ix}+e^{-ix})={\frac {1}{2}}(g_{1}+g_{-1}).}

One can create a positive definite function 
 
 
 
 f
 :
 X
 →
 
 C
 
 
 
 {\displaystyle f\colon X\to \mathbb {C} }
 
 easily from positive definite function 
 
 
 
 f
 :
 
 R
 
 →
 
 C
 
 
 
 {\displaystyle f\colon \mathbb {R} \to \mathbb {C} }
 
 for any <a href="/facts/Vector_space/M1pxMLx2">vector space</a> 
 
 
 
 X
 
 
 {\displaystyle X}
 
: choose a <a href="/facts/Linear_function/Ejvy8CKy">linear function</a> 
 
 
 
 ϕ
 :
 X
 →
 
 R
 
 
 
 {\displaystyle \phi \colon X\to \mathbb {R} }
 
 and define 
 
 
 
 
 f
 
 ∗
 
 
 :=
 f
 ∘
 ϕ
 
 
 {\displaystyle f^{*}:=f\circ \phi }
 
.
Then

u
          
            ∗
          
        
        
          A
          
            (
            
              f
              
                ∗
              
            
            )
          
        
        u
        =
        
          ∑
          
            j
            ,
            k
            =
            1
          
          
            n
          
        
        
          
            
              u
              
                k
              
            
            ¯
          
        
        
          u
          
            j
          
        
        
          f
          
            ∗
          
        
        (
        
          x
          
            k
          
        
        −
        
          x
          
            j
          
        
        )
        =
        
          ∑
          
            j
            ,
            k
            =
            1
          
          
            n
          
        
        
          
            
              u
              
                k
              
            
            ¯
          
        
        
          u
          
            j
          
        
        f
        (
        ϕ
        (
        
          x
          
            k
          
        
        )
        −
        ϕ
        (
        
          x
          
            j
          
        
        )
        )
        =
        
          u
          
            ∗
          
        
        
          
            
              
                A
                ~
              
            
          
          
            (
            f
            )
          
        
        u
        ≥
        0
        ,
      
    
    {\displaystyle u^{*}A^{(f^{*})}u=\sum _{j,k=1}^{n}{\overline {u_{k}}}u_{j}f^{*}(x_{k}-x_{j})=\sum _{j,k=1}^{n}{\overline {u_{k}}}u_{j}f(\phi (x_{k})-\phi (x_{j}))=u^{*}{\tilde {A}}^{(f)}u\geq 0,}

where 
 
 
 
 
 
 
 
 A
 ~
 
 
 
 
 (
 f
 )
 
 
 =
 
 
 (
 
 
 f
 (
 ϕ
 (
 
 x
 
 i
 
 
 )
 −
 ϕ
 (
 
 x
 
 j
 
 
 )
 )
 =
 f
 (
 
 
 
 
 x
 ~
 
 
 
 
 i
 
 
 −
 
 
 
 
 x
 ~
 
 
 
 
 j
 
 
 )
 
 
 
 )
 
 
 
 i
 ,
 j
 
 
 
 
 {\displaystyle {\tilde {A}}^{(f)}={\big (}f(\phi (x_{i})-\phi (x_{j}))=f({\tilde {x}}_{i}-{\tilde {x}}_{j}){\big )}_{i,j}}
 
 where 
 
 
 
 
 
 
 
 x
 ~
 
 
 
 
 k
 
 
 :=
 ϕ
 (
 
 x
 
 k
 
 
 )
 
 
 {\displaystyle {\tilde {x}}_{k}:=\phi (x_{k})}
 
 are distinct as 
 
 
 
 ϕ
 
 
 {\displaystyle \phi }
 
 is <a href="/facts/Linear/fChtg3KC">linear</a>.<a class="footnote-ref" id="fnref:1" href="#fn:1">1</a>

<h3>Bochner's theorem</h3>
Main article: <a href="/facts/Bochner%2527s_theorem/uQ97G6pP">Bochner's theorem</a>
Positive-definiteness arises naturally in the theory of the <a href="/facts/Fourier_transform/XXhKJtc8">Fourier transform</a>; it can be seen directly that to be positive-definite it is sufficient for f to be the Fourier transform of a function g on the real line with g(y) ≥ 0.
The converse result is <a href="/facts/Bochner%2527s_theorem/uQ97G6pP">Bochner's theorem</a>, stating that any <a href="/facts/Continuous_function/sKbl02pB">continuous</a> positive-definite function on the real line is the Fourier transform of a (positive) <a href="/facts/Measure_(mathematics)/yCq7zlI4">measure</a>.<a class="footnote-ref" id="fnref:2" href="#fn:2">2</a>

<h4>Applications</h4>
In <a href="/facts/Statistics/DNPsKGYU">statistics</a>, and especially <a href="/facts/Bayesian_statistics/9w7b1Bw4">Bayesian statistics</a>, the theorem is usually applied to real functions. Typically, n scalar measurements of some scalar value at points in 
 
 
 
 
 R
 
 d
 
 
 
 
 {\displaystyle R^{d}}
 
 are taken and points that are mutually close are required to have measurements that are highly correlated. In practice, one must be careful to ensure that the resulting covariance matrix (an n × n matrix) is always positive-definite. One strategy is to define a correlation matrix A which is then multiplied by a scalar to give a <a href="/facts/Covariance_matrix/kLhgjHC6">covariance matrix</a>: this must be positive-definite. Bochner's theorem states that if the correlation between two points is dependent only upon the distance between them (via function f), then function f must be positive-definite to ensure the covariance matrix A is positive-definite. See <a href="/facts/Kriging/Rk7X4qP5">Kriging</a>.
In this context, Fourier terminology is not normally used and instead it is stated that f(x) is the <a href="/facts/Characteristic_function_(probability_theory)/rl3uAGKM">characteristic function</a> of a <a href="/facts/Symmetric/hpu7DK8p">symmetric</a> <a href="/facts/Probability_density_function/zvfybna4">probability density function (PDF)</a>.

<h3>Generalization</h3>
Main article: <a href="/facts/Positive-definite_function_on_a_group/Vn71oAYE">Positive-definite function on a group</a>
One can define positive-definite functions on any <a href="/facts/Locally_compact_abelian_topological_group/fbD3d8kz">locally compact abelian topological group</a>; Bochner's theorem extends to this context. Positive-definite functions on groups occur naturally in the <a href="/facts/Representation_theory/3ydLtaGH">representation theory</a> of groups on <a href="/facts/Hilbert_space/aDFMh4lV">Hilbert spaces</a> (i.e. the theory of <a href="/facts/Unitary_representation/W1XRatG8">unitary representations</a>).

<h2 id="definition-2">Definition 2</h2>
Alternatively, a function 
 
 
 
 f
 :
 
 
 R
 
 
 n
 
 
 →
 
 R
 
 
 
 {\displaystyle f:\mathbb {R} ^{n}\to \mathbb {R} }
 
 is called positive-definite on a <a href="/facts/Neighborhood_(mathematics)/XHKgrpkp">neighborhood</a> D of the origin if 
 
 
 
 f
 (
 0
 )
 =
 0
 
 
 {\displaystyle f(0)=0}
 
 and 
 
 
 
 f
 (
 x
 )
 >
 0
 
 
 {\displaystyle f(x)>0}
 
 for every non-zero 
 
 
 
 x
 ∈
 D
 
 
 {\displaystyle x\in D}
 
.<a class="footnote-ref" id="fnref:3" href="#fn:3">3</a><a class="footnote-ref" id="fnref:4" href="#fn:4">4</a> 
Note that this definition conflicts with definition 1, given above.
In physics, the requirement that 
 
 
 
 f
 (
 0
 )
 =
 0
 
 
 {\displaystyle f(0)=0}
 
 is sometimes dropped (see, e.g., Corney and Olsen<a class="footnote-ref" id="fnref:5" href="#fn:5">5</a>).

<h2 id="see-also">See also</h2>
<ul><li><a href="/facts/Positive_definiteness/5Jvw9A5o">Positive definiteness</a></li>
<li><a href="/facts/Positive-definite_kernel/Ws3zkxfl">Positive-definite kernel</a></li></ul>

<ul><li>Christian Berg, Christensen, Paul Ressel. Harmonic Analysis on Semigroups, GTM, Springer Verlag.</li>
<li>Z. Sasvári, Positive Definite and Definitizable Functions, Akademie Verlag, 1994</li>
<li>Wells, J. H.; Williams, L. R. Embeddings and extensions in analysis. Ergebnisse der Mathematik und ihrer Grenzgebiete, Band 84. Springer-Verlag, New York-Heidelberg, 1975. vii+108 pp.</li></ul>
<h2 id="notes">Notes</h2>

<h2 id="external-links">External links</h2>
<ul><li><a href="https://www.encyclopediaofmath.org/index.php?title=Positive-definite_function">"Positive-definite function"</a>, <a href="/facts/Encyclopedia_of_Mathematics/WC6mGtPm">Encyclopedia of Mathematics</a>, <a href="/facts/European_Mathematical_Society/B3h7b672">EMS Press</a>, 2001 [1994]</li></ul>

<h2 id="references">References</h2>

<ol>
<li id="fn:1">Cheney, Elliot Ward (2009). A course in Approximation Theory. American Mathematical Society. pp. 77–78. ISBN 9780821847985. Retrieved 3 February 2022. <a href="9780821847985" target="_blank">9780821847985</a> <a href="#fnref:1" class="footnote-back-ref">↩</a></li>
<li id="fn:2">Bochner, Salomon (1959). Lectures on Fourier integrals. Princeton University Press. <a href="/wiki/Salomon_Bochner" target="_blank">/wiki/Salomon_Bochner</a> <a href="#fnref:2" class="footnote-back-ref">↩</a></li>
<li id="fn:3">Verhulst, Ferdinand (1996). Nonlinear Differential Equations and Dynamical Systems (2nd ed.). Springer. ISBN 3-540-60934-2. <a href="3-540-60934-2" target="_blank">3-540-60934-2</a> <a href="#fnref:3" class="footnote-back-ref">↩</a></li>
<li id="fn:4">Hahn, Wolfgang (1967). Stability of Motion. Springer. <a href="https://archive.org/details/stabilityofmotio0000hahn" target="_blank">https://archive.org/details/stabilityofmotio0000hahn</a> <a href="#fnref:4" class="footnote-back-ref">↩</a></li>
<li id="fn:5">Corney, J. F.; Olsen, M. K. (19 February 2015). "Non-Gaussian pure states and positive Wigner functions". Physical Review A. 91 (2): 023824. arXiv:1412.4868. Bibcode:2015PhRvA..91b3824C. doi:10.1103/PhysRevA.91.023824. ISSN 1050-2947. S2CID 119293595. <a href="/wiki/ArXiv_(identifier)" target="_blank">/wiki/ArXiv_(identifier)</a> <a href="#fnref:5" class="footnote-back-ref">↩</a></li>
</ol>

Positive-definite function open-in-new

Positive-definite function