Wirtingers Calculus in general Hilbert Spaces

Wirtinger's calculus [15] has become very popular in the signal processing community mainly in the context of complex adaptive filtering [13,7,1,2,12,8,4,10], as a means of computing, in an elegant way, gradients of real valued cost functions defined on complex domains (C ν ). Such functions, obviously, are not holomorphic and therefore the complex derivative cannot be used. Instead, if we consider that the cost function is defined on a Euclidean domain with a double dimensionality (R 2ν ), then the real derivatives may be employed. The price of this approach is that the computations become cumbersome and tedious. Wirtinger's calculus provides an alternative equivalent formulation, that is based on simple rules and principles and which bears a great resemblance to the rules of the standard complex derivative. Although this methodology is relatively known in the German speaking countries and has been applied to practical applications [3,5], only recently has become popular in the signal processing community, mostly due to the works of Picinbono on widely linear estimation filters [13].

Most Complex Analysis’ textbooks deal with complex analytic (i.e., holomorphic) functions and their properties, which in order to be studied properly a great deal of notions from topology, differential geometry, calculus on manifolds and from other mathematical fields need to be employed. Therefore, most of these materials are accessible only to the specialist. It is only natural that in the scope of this literature, Wirtinger’s calculus is usually ignored, since it involves non-holomorphic functions. Nevertheless, in some of these textbooks, the ideas of Wirtinger’s calculus are mentioned, although, in most cases, they are presented either superficially, or they are introduced in a completely different set-up (mainly as a bi-product of the Cauchy Riemann conditions). Some of these textbooks are [14,11,9]. However, most of these works are highly specialized and technically abstruse, and therefore not recommended for someone who wants only to understand the concepts of Wirtinger’s calculus and use them in his/her field. Moreover, a rigorous and self-consistent presentation of the main ideas of Wirtinger’s calculus cannot be found in any of those works. An excellent and highly recommended first attempt to summarize all the related concepts and present them in a unified framework is the introductory text of K. Kreutz-Delgado [6]. The aim of the present manuscript is twofold: a) it endeavors to provide a more rigorous presentation of the related material, focusing on aspects that the author finds more insightful and b) it extends the notions of Wirtinger’s calculus on general Hilbert spaces (such as Reproducing Hilbert Kernel Spaces).

A common misconception (usually done by beginners in the field) is that Wirtinger’s calculus uses an alternative definition of derivatives and therefore results in different gradient rules in minimization problems. We should emphasize that the theoretical foundation of Wirtinger’s calculus is the common definition of the real derivative. However, it turns out that when the complex structure is taken into account, the real derivatives may be described using an equivalent and more elegant formulation which bears a surprising resemblance with the complex derivative. Therefore, simple rules may be derived and the computations of the gradients, which may become tedious if the double dimensional space R 2ν is considered, are simplified.

The manuscript has two main parts. Section 2, deals with ordinary Wirtinger’s calculus for functions of one complex variable, while in section 3 the main notions and results of the extended Wirtinger’s Calculus in RKHSs are presented. Throughout the paper, we will denote the set of all integers, real and complex numbers by N, R and C respectively. Vector or matrix valued quantities appear in boldfaced symbols.

The present report, has been inspired by the need of the author and its colleagues to understand the underlying theory of Wirtinger’s Calculus and to further extend it to include the kernel case. Many parts have been considerably improved after long discussions with prof. S. Theodoridis and L. Dalla.

Consider the function f : X ⊆ C → C, f (z) = f (x + iy) = u(x, y) + v(x, y)i, where u, and v are real valued functions defined on an open subset X of R 2 . Any such function, f , may be regarded as defined either on a subset of C or on a subset of R 2 . Furthermore, f may be regarded either as a complex valued function, or as a vector valued function, which takes values in R 2 . Therefore, we may equivalently write: f (z) = f (x + iy) = f (x, y) = u(x, y) + iv(x, y) = (u(x, y), v(x, y)).

The complex derivative of f at c, if it exists, is defined as the limit:

This definition, although similar with the typical real derivative of elementary calculus, exploits the complex structure of X. More specifically, the division that appears in the definition is based on the co

…(Full text truncated)…

This content is AI-processed based on ArXiv data.