Difference between revisions of "Discrepancy"
(Importing text file) |
|||
(3 intermediate revisions by 2 users not shown) | |||
Line 1: | Line 1: | ||
− | ''of a sequence | + | ''of a sequence $\omega=(\mathbf{x}_1,\ldots,\mathbf{x}_N)$ of points from the unit $s$-dimensional cube $K_s = \{\mathbf{x} : 0 \le x_\nu < 1\,,\ \nu=1,\ldots,s \}$'' |
The norm of the functional | The norm of the functional | ||
+ | \begin{equation}\label{eq:1} | ||
+ | \phi(\alpha;\omega) = |V| - \frac{N(V)}{N}\,, | ||
+ | \end{equation} | ||
+ | calculated in some metric. Here, $|V|$ and $N(V)$ are, respectively, the volume of the domain $V = \{\mathbf{x} : 0 \le x_\nu < \alpha_\nu\,,\ \nu=1,\ldots,s \}$ and the number of the points of $\omega$ belonging to $V$. If one considers the distribution of the points of $\omega$ over domains of the type $V = \{\mathbf{x} : \alpha_\nu \le x_\nu < \beta_\nu\,,\ \nu=1,\ldots,s \}$, then, in formula (1), $\phi(\alpha;\omega)$ is usually replaced by $\phi(\alpha,\beta;\omega)$. | ||
− | + | The following norms of the functional \eqref{eq:1} are most often used: | |
+ | $$ | ||
+ | D_N(\omega) = \sup_{\alpha,\beta\in K_s} |\phi(\alpha,\beta;\omega)|\ , | ||
+ | $$ | ||
+ | $$ | ||
+ | D_N^*(\omega) = \sup_{\alpha\in K_s} |\phi(\alpha;\omega)|\ , | ||
+ | $$ | ||
+ | $$ | ||
+ | D_N(\omega,L_p) = \left({ \int_0^1\cdots\int_0^1 |\phi(\alpha;\omega)|^p d\alpha_1\ldots d\alpha_s }\right)^{1/p} \ . | ||
+ | $$ | ||
− | + | A sequence $\omega=(\mathbf{x}_1,\ldots,\mathbf{x}_N,\ldots)$ of points from the $s$-dimensional unit cube $K_s$ is ''uniformly distributed'' if and only if [[#References|[1]]] | |
+ | $$ | ||
+ | \lim_{N\rightarrow\infty} D_N(\omega) = 0 \ . | ||
+ | $$ | ||
− | + | For any infinite sequence $\omega=(x_1,\ldots,x_N,\ldots)$ of one-dimensional points the following theorem [[#References|[3]]] is valid: | |
− | + | $$ | |
− | + | \limsup N D_N(\omega) = \infty \ . | |
− | + | $$ | |
− | + | For any such sequence $\omega$ it is possible to find a sequence $N_1,\ldots,N_k,\ldots$ such that for $N = N_k$ one has [[#References|[4]]], | |
− | + | $$ | |
− | + | N D_N(\omega) > C_1 \sqrt{\log N} \ . | |
− | + | $$ | |
− | + | The final result [[#References|[5]]] for infinite sequences of one-dimensional points is that for $N = N_k$: | |
− | + | $$ | |
− | + | N D_N(\omega) > C_2 \log N \ . | |
− | + | $$ | |
− | For any infinite sequence | ||
− | |||
− | |||
− | |||
− | For any sequence | ||
− | |||
− | |||
− | |||
− | The final result [[#References|[5]]] for infinite sequences of one-dimensional points is that for | ||
− | |||
− | |||
Studies were made of the discrepancies of various concrete sequences [[#References|[6]]]–[[#References|[8]]], and the estimates from above | Studies were made of the discrepancies of various concrete sequences [[#References|[6]]]–[[#References|[8]]], and the estimates from above | ||
+ | $$ | ||
+ | N D_N(\omega,L_2) \le C_3(s) \log^{s+1} N \ , | ||
+ | $$ | ||
+ | $$ | ||
+ | N D_N(\omega) \le C_4(s) \log^s N | ||
+ | $$ | ||
+ | were obtained, respectively, for finite and infinite sequences, as well as an estimate from below [[#References|[4]]]: For any sequence of $N$ points, the following inequality is valid: | ||
+ | $$ | ||
+ | N D_N(\omega,L_2) \ge C_5(s) \log^{(s+1)/2} N \ . | ||
+ | $$ | ||
− | + | For any infinite sequence $\omega = \{\mathbf{x}_n \in K_s \}$ it is possible to find a sequence of numbers $N_1,\ldots,N_k,\ldots$ such that for $N = N_k$ one has | |
− | + | $$ | |
− | + | N D_N(\omega,L_2) \ge C_6(s) \log^{s/2} N \ . | |
− | + | $$ | |
− | |||
− | |||
− | |||
− | |||
− | For any infinite sequence | ||
− | |||
− | |||
Also, | Also, | ||
− | + | $$ | |
− | + | D_N(\omega) \ge D_N(\omega,L_2) \ . | |
− | + | $$ | |
− | |||
− | |||
− | |||
====Comments==== | ====Comments==== | ||
− | See also [[ | + | See also [[Distribution modulo one]]; [[Distribution modulo one, higher-dimensional]]; [[Uniform distribution]]. |
====References==== | ====References==== | ||
− | <table><TR><TD valign="top">[a1]</TD> <TD valign="top"> J. Beck, W.L. Chen, "Irregularities of distribution" , Cambridge Univ. Press (1987)</TD></TR></table> | + | <table> |
+ | <TR><TD valign="top">[1]</TD> <TD valign="top"> H. Weyl, "Ueber die Gleichverteilung von Zahlen mod Eins" ''Math. Ann.'' , '''77''' (1916) pp. 313–352</TD></TR> | ||
+ | <TR><TD valign="top">[2]</TD> <TD valign="top"> J.G. van der Corput, "Verteilungsfunktionen" ''Proc. Koninkl. Ned. Akad. Wet. A'' , '''38''' : 8 (1935) pp. 813–821; 1058–1066</TD></TR> | ||
+ | <TR><TD valign="top">[3]</TD> <TD valign="top"> T. van Aardenne-Ehrenfest, "On the impossibility of a just distribution" ''Indag. Math.'' , '''11''' (1949) pp. 264–269</TD></TR> | ||
+ | <TR><TD valign="top">[4]</TD> <TD valign="top"> K.F. Roth, "On irregularities of distribution" ''Mathematika'' , '''1''' (1954) pp. 73–79 {{ZBL|0057.28604}}</TD></TR> | ||
+ | <TR><TD valign="top">[5]</TD> <TD valign="top"> W.M. Schmidt, "Irregularities of distribution VII" ''Acta Arithm.'' , '''21''' (1972) pp. 45–50</TD></TR> | ||
+ | <TR><TD valign="top">[6]</TD> <TD valign="top"> J.H. Halton, "On the efficiency of certain quasi-random sequences of points in evaluating multi-dimensional integrals" ''Numer. Math.'' , '''2''' : 2 (1960) pp. 84–90</TD></TR> | ||
+ | <TR><TD valign="top">[7]</TD> <TD valign="top"> I.M. Sobol', "The distribution of points in a cube and the approximate evaluation of integrals" ''USSR Comp. Math. and Math. Phys.'' , '''7''' : 4 (1967) pp. 86–112 ''Zh. Vychisl. Mat. i Mat. Fiz.'' , '''7''' : 4 (1967) pp. 784–802</TD></TR> | ||
+ | <TR><TD valign="top">[8]</TD> <TD valign="top"> N.M. Korobov, "Number-theoretical methods in approximate analysis" , Moscow (1963) (In Russian)</TD></TR> | ||
+ | <TR><TD valign="top">[9]</TD> <TD valign="top"> L. Kuipers, H. Niederreiter, "Uniform distribution of sequences" , Wiley (1974) {{ZBL|0281.10001}}; repr. Dover (2006) {{ISBN|0-486-45019-8}} </TD></TR> | ||
+ | <TR><TD valign="top">[a1]</TD> <TD valign="top"> J. Beck, W.L. Chen, "Irregularities of distribution" , Cambridge Univ. Press (1987) {{ZBL|0617.10039}}</TD></TR> | ||
+ | </table> | ||
+ | |||
+ | {{TEX|done}} |
Latest revision as of 15:35, 8 October 2023
of a sequence $\omega=(\mathbf{x}_1,\ldots,\mathbf{x}_N)$ of points from the unit $s$-dimensional cube $K_s = \{\mathbf{x} : 0 \le x_\nu < 1\,,\ \nu=1,\ldots,s \}$
The norm of the functional \begin{equation}\label{eq:1} \phi(\alpha;\omega) = |V| - \frac{N(V)}{N}\,, \end{equation} calculated in some metric. Here, $|V|$ and $N(V)$ are, respectively, the volume of the domain $V = \{\mathbf{x} : 0 \le x_\nu < \alpha_\nu\,,\ \nu=1,\ldots,s \}$ and the number of the points of $\omega$ belonging to $V$. If one considers the distribution of the points of $\omega$ over domains of the type $V = \{\mathbf{x} : \alpha_\nu \le x_\nu < \beta_\nu\,,\ \nu=1,\ldots,s \}$, then, in formula (1), $\phi(\alpha;\omega)$ is usually replaced by $\phi(\alpha,\beta;\omega)$.
The following norms of the functional \eqref{eq:1} are most often used: $$ D_N(\omega) = \sup_{\alpha,\beta\in K_s} |\phi(\alpha,\beta;\omega)|\ , $$ $$ D_N^*(\omega) = \sup_{\alpha\in K_s} |\phi(\alpha;\omega)|\ , $$ $$ D_N(\omega,L_p) = \left({ \int_0^1\cdots\int_0^1 |\phi(\alpha;\omega)|^p d\alpha_1\ldots d\alpha_s }\right)^{1/p} \ . $$
A sequence $\omega=(\mathbf{x}_1,\ldots,\mathbf{x}_N,\ldots)$ of points from the $s$-dimensional unit cube $K_s$ is uniformly distributed if and only if [1] $$ \lim_{N\rightarrow\infty} D_N(\omega) = 0 \ . $$
For any infinite sequence $\omega=(x_1,\ldots,x_N,\ldots)$ of one-dimensional points the following theorem [3] is valid: $$ \limsup N D_N(\omega) = \infty \ . $$ For any such sequence $\omega$ it is possible to find a sequence $N_1,\ldots,N_k,\ldots$ such that for $N = N_k$ one has [4], $$ N D_N(\omega) > C_1 \sqrt{\log N} \ . $$ The final result [5] for infinite sequences of one-dimensional points is that for $N = N_k$: $$ N D_N(\omega) > C_2 \log N \ . $$
Studies were made of the discrepancies of various concrete sequences [6]–[8], and the estimates from above $$ N D_N(\omega,L_2) \le C_3(s) \log^{s+1} N \ , $$ $$ N D_N(\omega) \le C_4(s) \log^s N $$ were obtained, respectively, for finite and infinite sequences, as well as an estimate from below [4]: For any sequence of $N$ points, the following inequality is valid: $$ N D_N(\omega,L_2) \ge C_5(s) \log^{(s+1)/2} N \ . $$
For any infinite sequence $\omega = \{\mathbf{x}_n \in K_s \}$ it is possible to find a sequence of numbers $N_1,\ldots,N_k,\ldots$ such that for $N = N_k$ one has $$ N D_N(\omega,L_2) \ge C_6(s) \log^{s/2} N \ . $$
Also, $$ D_N(\omega) \ge D_N(\omega,L_2) \ . $$
Comments
See also Distribution modulo one; Distribution modulo one, higher-dimensional; Uniform distribution.
References
[1] | H. Weyl, "Ueber die Gleichverteilung von Zahlen mod Eins" Math. Ann. , 77 (1916) pp. 313–352 |
[2] | J.G. van der Corput, "Verteilungsfunktionen" Proc. Koninkl. Ned. Akad. Wet. A , 38 : 8 (1935) pp. 813–821; 1058–1066 |
[3] | T. van Aardenne-Ehrenfest, "On the impossibility of a just distribution" Indag. Math. , 11 (1949) pp. 264–269 |
[4] | K.F. Roth, "On irregularities of distribution" Mathematika , 1 (1954) pp. 73–79 Zbl 0057.28604 |
[5] | W.M. Schmidt, "Irregularities of distribution VII" Acta Arithm. , 21 (1972) pp. 45–50 |
[6] | J.H. Halton, "On the efficiency of certain quasi-random sequences of points in evaluating multi-dimensional integrals" Numer. Math. , 2 : 2 (1960) pp. 84–90 |
[7] | I.M. Sobol', "The distribution of points in a cube and the approximate evaluation of integrals" USSR Comp. Math. and Math. Phys. , 7 : 4 (1967) pp. 86–112 Zh. Vychisl. Mat. i Mat. Fiz. , 7 : 4 (1967) pp. 784–802 |
[8] | N.M. Korobov, "Number-theoretical methods in approximate analysis" , Moscow (1963) (In Russian) |
[9] | L. Kuipers, H. Niederreiter, "Uniform distribution of sequences" , Wiley (1974) Zbl 0281.10001; repr. Dover (2006) ISBN 0-486-45019-8 |
[a1] | J. Beck, W.L. Chen, "Irregularities of distribution" , Cambridge Univ. Press (1987) Zbl 0617.10039 |
Discrepancy. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Discrepancy&oldid=14906