Some Classes of Random Fields inn-Dimensional Space, Related to Stationary Random Processes
Author:
- A. M. YaglomAYA. M. Yaglom
- 1 January 1957
- journal article
- Published by Society for Industrial & Applied Mathematics (SIAM) in Theory of Probability & Its Applications
- Vol. 2 (3) , 273-320
Abstract
The purpose of this paper is to establish a spectral theory for certain types of random fields and random generalized fields (multidimensional random distributions) in the Euclidean n-space $R_n $ similar to the well-known spectral theory for stationary random processes. Let D denote the Schwartz space of all complex-valued $C_\infty $ functions $\varphi ({\bf x})$ defined on $R_n $ whose carrier is compact. Following Ito [6] and Gelfand [7] we shall call the random linear functional $\xi (\varphi )$ on D satisfying (1.4) the random generalized field. We can identify a continuous random field $\xi ({\bf x})$ on $R_n $ with a random generalized field (1.5) and therefore we can consider the ordinary random fields $\xi ({\bf x})$ as special cases of random generalized fields. We shall only deal with the first moment $m(\varphi )$ and the second moment $B(\varphi _1 ,\varphi _2 )$ of the random generalized field $\xi (\varphi )$ and shall call them a mean value functional and a covariance functional of this field. For an ordinary random field the mean value and covariance function defined by (1.1) play a similar role. The random generalized field $\xi (\varphi )$ is called homogeneous if its mean value functional $m(\varphi )$ and covariance functional $B(\varphi _1 ,\varphi _2 )$ are invariant under all shift transformations in the space D, i.e. if the relations (2.2) and (2.3) hold true. The theory of this homogeneous random generalized field, which is very similar to Ito’s theory of a stationary random distribution [6], is treated in Section 2. The main results of the section deal with the spectral representation of the covariance functional of such fields and of random generalized fields themselves. Let $D_1 $ denote the subspace of the space D consisting of all functions $\varphi (x)$ satisfying (3.1). The continuous linear functional $\xi (\varphi )$ on $D_1 $ is called a locally homogeneous random generalized field if its mean value functional $m(\varphi )$ and its covariance functional $B(\varphi _1 ,\varphi _2 )$ are invariant under all shift transformations in $D_1 $. The locally homogeneous random fields mentioned above are treated in Section 3. In this section we obtain the spectral representation of a covariance functional of the locally homogeneous random generalized field and the spectral representation of this field itself; these results generalize the spectral theory for random processes with stationary increments. The homogeneous random generalized field is called homogeneous and isotropic if its mean value functional $m(\varphi )$ and covariance functional $B(\varphi _1 ,\varphi _2 )$ are invariant under 320 Some Classes of Random Fields all transformations in D induced by orthogonal transformations (motions and reflections) in the space $R_n $. In the case of the n-dimensional random generalized field ${\bf \xi} (\varphi ) = = \{ \xi _1 (\varphi ), \cdots ,\xi _n (\varphi )\} $ its mean value functional ${\bf m}(\varphi ) = \{ m_1 (\varphi ), \cdots ,m_n (\varphi )\} $ forms a vector in $R_n $ and its covariance functionals $B_{ij} (\varphi _1 ,\varphi _2 )$ form a tensor in $R_n $. The n-dimensional homogeneous field ${\bf \xi} (\varphi )$ is called a homogeneous and isotropic random generalized vector field if the vector and the tensor are invariant under all motions and reflections in $R_n $ and the simultaneous transformations in D induced by this motion or reflection Homogeneous and isotropic random generalized fields and random generalized vector fields are treated in Section 4. In this section we obtain the general form of the functionals $m(\varphi )$ (or ${\bf m}(\varphi )$) and $B(\varphi _1 ,\varphi _2 )$ (or $B_{ij} (\varphi _1 ,\varphi _2 )$) for these fields. The locally homogeneous random generalized field $\xi (\varphi )$, $\varphi \in D_1 $ is called locally homogeneous and locally isotropic if its mean value functional $m(\varphi )$ and covariance functional $B(\varphi _1 ,\varphi _2 )$ are invariant under all transformations in $D_1 $ induced by orthogonal transformations in $R_n $; it is also clear how we can define the notion of the locally homogeneous and locally isotropic random generalized vector field ${\bf \xi} (\varphi )$, $\varphi \in D_1 $. Locally homogeneous and locally isotropic random generalized vector fields are treated in Section 5. Here we obtain the general form of the mean value functional and of the covariance for these fields; in particular, we obtain the general form of a mean value and of a covariance function of ordinary random fields (scalar and vector) which are locally homogeneous and locally isotropic in the sense of Kolmogorov [17].
Keywords
This publication has 8 references indexed in Scilit:
- Stationary random distributionsKyoto Journal of Mathematics, 1954
- The spectra of turbulence in a compressible fluid; eddy turbulence and random noiseMathematical Proceedings of the Cambridge Philosophical Society, 1952
- The Classical Groups1947
- Kolmogoroff's theory of locally isotropic turbulenceMathematical Proceedings of the Cambridge Philosophical Society, 1947
- Fourier integrals and metric geometryTransactions of the American Mathematical Society, 1941
- Metric Spaces and Completely Monotone FunctionsAnnals of Mathematics, 1938
- On the Statistical Theory of Isotropic TurbulenceProceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1938
- Statistical theory of turbulencProceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1935