BZPEERReading, trust and collaboration for researchDiscover
Workspaces
Network
Opportunities
Account
Researcher profile
Palle E. T. Jorgensen contributes to research discovery and scholarly infrastructure.
Trust snapshot
Actions
Identity and collaboration
Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.
Log in to claimDirect collaboration
Claim this author entity first to unlock direct invitations.
Research graph
Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.
BZPEER is loading the nearby papers, people, topics and institutions for this page.
Published work
In this paper we show how specific families of positive definite kernels serve as powerful tools in analyses of iteration algorithms for multiple layer feedforward Neural Network models. Our focus is on particular kernels that adapt well to learning algorithms for data-sets/features which display intrinsic self-similarities at feedforward iterations of scaling.
We study non-linear data-dimension reduction. We are motivated by the classical linear framework of Principal Component Analysis. In nonlinear case, we introduce instead a new kernel-Principal Component Analysis, manifold and feature space transforms. Our results extend earlier work for probabilistic Karhunen-Loève transforms on compression of wavelet images. Our object is algorithms for optimization, selection of efficient bases, or components, which serve to minimize entropy and error; and hence to improve digital representation of images, and hence of optimal storage, and transmission. We prove several new theorems for data-dimension reduction. Moreover, with the use of frames in Hilbert space, and a new Hilbert-Schmidt analysis, we identify when a choice of Gaussian kernel is optimal.
We show that a Krein-Feller operator is naturally associated to a fixed measure $μ$, assumed positive, $σ$-finite, and non-atomic. Dual pairs of operators are introduced, carried by the two Hilbert spaces, $L^{2}\left(μ\right)$ and $L^{2}\left(λ\right)$, where $λ$ denotes Lebesgue measure. An associated operator pair consists of two specific densely defined (unbounded) operators, each one contained in the adjoint of the other. This then yields a rigorous analysis of the corresponding $μ$-Krein-Feller operator as a closable quadratic form. As an application, for a given measure $μ$, including the case of fractal measures, we compute the associated diffusion, semigroup, Dirichlet forms, and $μ$-generalized heat equation.
The past decade has seen a flourishing of advances in harmonic analysis of graphs. They lie at the crossroads of graph theory and such analytical tools as graph Laplacians, Markov processes and associated boundaries, analysis of path-space, harmonic analysis, dynamics, and tail-invariant measures. Motivated by recent advances for the special case of Bratteli diagrams, our present focus will be on those graph systems $G$ with the property that the sets of vertices $V$ and edges $E$ admit discrete level structures. A choice of discrete levels in turn leads to new and intriguing discrete-time random-walk models. Our main extension (which greatly expands the earlier analysis of Bratteli diagrams) is the case when the levels in the graph system $G$ under consideration are now allowed to be standard measure spaces. Hence, in the measure framework, we must deal with systems of transition probabilities, as opposed to incidence matrices (for the traditional Bratteli diagrams). The paper is divided into two parts, (i) the special case when the levels are countable discrete systems, and (ii) the (non-atomic) measurable category, i.e., when each level is a prescribed measure space with standard Borel structure. The study of the two cases together is motivated in part by recent new results on graph-limits. Our results depend on a new analysis of certain duality systems for operators in Hilbert space; specifically, one dual system of operator for each level. We prove new results in both cases, (i) and (ii); and we further stress both similarities, and differences, between results and techniques involved in the two cases.
We find conditions under which two measure preserving actions of two groups on the same space have a common fundamental domain. Our results apply to commuting actions with separate fundamental domains, lattices in groups of polynomial growth, and some semidirect products. We prove that two lattices of equal co-volume in a group of polynomial growth, one acting on the left, the other on the right, have a common fundamental domain.
We show that the spectral-tile implication in the Fuglede conjecture in dimension 1 is equivalent to a Universal Tiling Conjecture and also to similar forms of the same implication for some simpler sets, such as unions of intervals with rational or integer endpoints.
We study a class of dynamical systems in $L^2$ spaces of infinite products $X$. Fix a compact Hausdorff space $B$. Our setting encompasses such cases when the dynamics on $X = B^\bn$ is determined by the one-sided shift in $X$, and by a given transition-operator $R$. Our results apply to any positive operator $R$ in $C(B)$ such that $R1 = 1$. From this we obtain induced measures $Σ$ on $X$, and we study spectral theory in the associated $L^2(X,Σ)$. For the second class of dynamics, we introduce a fixed endomorphism $r$ in the base space $B$, and specialize to the induced solenoid $\Sol(r)$. The solenoid $\Sol(r)$ is then naturally embedded in $X = B^\bn$, and $r$ induces an automorphism in $\Sol(r)$. The induced systems will then live in $L^2(\Sol(r), Σ)$. The applications include wavelet analysis, both in the classical setting of $\br^n$, and Cantor-wavelets in the setting of fractals induced by affine iterated function systems (IFS). But our solenoid analysis includes such hyperbolic systems as the Smale-Williams attractor, with the endomorphism $r$ there prescribed to preserve a foliation by meridional disks. And our setting includes the study of Julia set-attractors in complex dynamics.
We study the boundary theory of a connected weighted graph $G$ from the viewpoint of stochastic integration. For the Hilbert space \HE of Dirichlet-finite functions on $G$, we construct a Gel'fand triple $S \ci {\mathcal H}_{\mathcal E} \ci S'$. This yields a probability measure $\mathbb{P}$ on $S'$ and an isometric embedding of ${\mathcal H}_{\mathcal E}$ into $L^2(S',\mathbb{P})$, and hence gives a concrete representation of the boundary as a certain class of "distributions" in $S'$. In a previous paper, we proved a discrete Gauss-Green identity for infinite networks which produces a boundary representation for harmonic functions of finite energy, given as a certain limit. In this paper, we use techniques from stochastic integration to make the boundary $\operatorname{bd}G$ precise as a measure space, and obtain a boundary integral representation as an integral over $S'$.
We examine the operator $U_5$ defined on $L^2(μ_{\frac14})$ where $μ_{\frac14}$ is the 1/4 Cantor measure. The operator $U_5$ scales the elements of the canonical exponential spectrum for $L^2(μ_{\frac14})$ by 5 --- that is, $Ue_γ = e_{5γ}$ where $e_γ(t) = e^{2πi γt}$. It is known that $U_5$ has a self-similar structure, which makes its spectrum, which is currently unknown, of particular interest. In order to better understand the spectrum of $U_5$, we demonstrate a decomposition of the projection valued measures and scalar spectral measures associated with $U_5$. We are also able to compute associated Radon-Nikodym derivatives between the scalar measures. Our decomposition utilizes a system of operators which form a representation of the Cuntz algebra $\mathcal{O}_2$.
Each Cantor measure (μ) with scaling factor 1/(2n) has at least one associated orthonormal basis of exponential functions (ONB) for L^2(μ). In the particular case where the scaling constant for the Cantor measure is 1/4 and two specific ONBs are selected for L^2(μ), there is a unitary operator U defined by mapping one ONB to the other. This paper focuses on the case in which one ONB (Γ) is the original Jorgensen-Pedersen ONB for the Cantor measure (μ) and the other ONB is is 5Γ. The main theorem of the paper states that the corresponding operator U is ergodic in the sense that only the constant functions are fixed by U.
Let $(G,c)$ be an infinite network, and let $\mathcal{E}$ be the canonical energy form. Let $Δ_2$ be the Laplace operator with dense domain in $\ell^2(G)$ and let $Δ_{\mathcal{E}}$ be the Laplace operator with dense domain in the Hilbert space $\mathcal{H}_\mathcal{E}$ of finite energy functions on $G$. It is known that $Δ_2$ is essentially self-adjoint, but that $Δ_{\mathcal{E}}$ is \emph{not}. In this paper, we characterize the Friedrichs extension of $Δ_{\mathcal{E}}$ in terms of $Δ_2$ and show that the spectral measures of the two operators are mutually absolutely continuous with Radon-Nikodym derivative $λ$ (the spectral parameter), in the complement of $λ=0$. We also give applications to the effective resistance on $(G,c)$. For transient networks, the Dirac measure at $λ= 0$ contributes to the spectral resolution of the Friedrichs extension of $Δ_{\mathcal{E}}$ but not to that of the self-adjoint $\ell^2$ Laplacian.
We present a characterization of the sets that appear as Fourier spectra of measures in terms of the existence of a strongly continuous representation of the ambient group that has a wandering vector for the given set.
Certain Bernoulli convolution measures (μ) are known to be spectral. Recently, much work has concentrated on determining conditions under which orthonormal Fourier bases (i.e. spectral bases) exist. For a fixed measure known to be spectral, the orthonormal basis need not be unique; indeed, there are often families of such spectral bases. Let λ= 1/(2n) for a natural number n and consider the Bernoulli measure (μ) with scale factor λ. It is known that L^2(μ) has a Fourier basis. We first show that there are Cuntz operators acting on this Hilbert space which create an orthogonal decomposition, thereby offering powerful algorithms for computations for Fourier expansions. When L^2(μ) has more than one Fourier basis, there are natural unitary operators U, indexed by a subset of odd scaling factors p; each U is defined by mapping one ONB to another. We show that the unitary operator U can also be orthogonally decomposed according to the Cuntz relations. Moreover, this operator-fractal U exhibits its own self-similarity.
We study a decomposition problem for a class of unitary representations associated with wavelet analysis, wavelet representations, but our framework is wider and has applications to multi-scale expansions arising in dynamical systems theory for non-invertible endomorphisms. Our main results offer a direct integral decomposition for the general wavelet representation, and we solve a question posed by Judith Packer. This entails a direct integral decomposition of the general wavelet representation. We further give a detailed analysis of the measures contributing to the decomposition into irreducible representations. We prove results for associated Martin boundaries, relevant for the understanding of wavelet filters and induced random-walks, as well as classes of harmonic functions. Our setting entails representations built from certain finite-to-one endomorphisms $r$ in compact metric spaces $X$, and we study their dilations to automorphisms in induced solenoids. Our wavelet representations are covariant systems formed from the dilated automorphisms. They depend on assigned measures $μ$ on $X$. It is known that when the data $(X, r, μ)$ are given the associated wavelet representation is typically reducible. By introducing wavelet filters associated to $(X, r)$ we build random walks in $X$, path-space measures, harmonic functions, and an associated Martin boundary. We construct measures on the solenoid $(X_\infty, r_\infty)$, built from $(X, r)$. We show that $r_\infty$ induces unitary operators $U$ on Hilbert space $\H$ and representations $π$ of the algebra $L^\infty(X)$ such that the pair $(U, r_\infty)$, together with the corresponding representation $π$ forms a cross-product in the sense of $C^*$-algebras. We note that the traditional wavelet representations fall within this wider framework of $(\H, U, π)$ covariant crossed products.
For a family of fractal measures, we find an explicit Fourier duality. The measures in the pair have compact support in $\br^d$, and they both have the same matrix scaling. But the two use different translation vectors, one by a subset $B$ in $\br^d$, and the other by a related subset $L$. Among other things, we show that there is then a pair of infinite discrete sets $Γ(L)$ and $Γ(B)$ in $\br^d$ such that the $Γ(L)$-Fourier exponentials are orthogonal in $L^2(μ_B)$, and the $Γ(B)$-Fourier exponentials are orthogonal in $L^2(μ_L)$. These sets of orthogonal "frequencies" are typically lacunary, and they will be obtained by scaling in the large. The nature of our duality is explored below both in higher dimensions and for examples on the real line. Our duality pairs do not always yield orthonormal Fourier bases in the respective $L^2(μ)$-Hilbert spaces, but depending on the geometry of certain finite orbits, we show that they do in some cases. We further show that there are new and surprising scaling symmetries of relevance for the ergodic theory of these affine fractal measures.
In recent papers a number of authors have considered Borel probability measures $μ$ in $\br^d$ such that the Hilbert space $L^2(μ)$ has a Fourier basis (orthogonal) of complex exponentials. If $μ$ satisfies this property, the set of frequencies in this set are called a spectrum for $μ$. Here we fix a spectrum, say $Γ$, and we study the possibilities for measures $μ$ having $Γ$ as spectrum.
We study the momentum operator defined on the disjoint union of two intervals. Even in one dimension, the question of two non-empty open and non-overlapping intervals has not been worked out in a way that extends the cases of a single interval and gives a list of the selfadjoint extensions. Starting with zero boundary conditions at the four endpoints, we characterize the selfadjoint extensions and undertake a systematic and complete study of the spectral theory of the selfadjoint extensions. In an application of our extension theory to harmonic analysis, we offer a new family of spectral pairs. Compared to earlier studies, it yields a more direct link between spectrum and geometry.
Motivated by potential theory on discrete spaces, we study a family of unbounded Hermitian operators in Hilbert space which generalize the usual graph-theoretic discrete Laplacian. These operators are discrete analogues of the classical conformal Laplacians and Hamiltonians from statistical mechanics. For an infinite discrete set $X$, we consider operators acting on Hilbert spaces of functions on $X$, and their representations as infinite matrices; the focus is on $\ell^2(X)$, and the energy space $\mathcal{H}_{\mathcal E}$. In particular, we prove that these operators are always essentially self-adjoint on $\ell^2(X)$, but may fail to be essentially self-adjoint on $\mathcal{H}_{\mathcal E}$. In the general case, we examine the von Neumann deficiency indices of these operators and explore their relevance in mathematical physics. Finally we study the spectra of the $\mathcal{H}_{\mathcal E}$ operators with the use of a new approximation scheme.
A resistance network is a connected graph $(G,c)$. The conductance function $c_{xy}$ weights the edges, which are then interpreted as resistors of possibly varying strengths. The relationship between the natural Dirichlet form $\mathcal E$ and the discrete Laplace operator $Δ$ on a finite network is given by $\mathcal E(u,v) = \la u, \Lap v\ra_2$, where the latter is the usual $\ell^2$ inner product. We extend this formula to infinite networks, where a new (boundary) term appears. The Laplace operator is typically unbounded in this context; we construct a reproducing kernel for the space of functions of finite energy which allows us to specify a dense domain for $Δ$ and give several criteria for the transience of the random walk on the network. The extended Gauss-Green identity and the reproducing kernel are the foundation for a boundary integral representation for harmonic functions of finite energy, akin to that of Martin boundary theory.
We introduce the notion of boundary representation for fractal Fourier expansions, starting with a familiar notion of spectral pairs for affine fractal measures. Specializing to one dimension, we establish boundary representations for these fractals. We prove that as sets these fractals arise as boundaries of functions in closed subspaces of the Hardy space $H^2$. By this we mean that there are lacunary subsets $Γ$ of the non-negative integers, and associated closed $Γ$-subspace in the Hardy space $H^2(\bd)$, $\bd$ denoting the disk, such that for every function $f$ in in $H^2(Γ)$, and for every point $z$ in $\bd$, $f(z)$ admits a boundary integral represented by an associated measure $μ$, with integration over $\suppμ$ placed as a Cantor subset on the circle $\bt := \{bd}(\bd)$. We study families of pairs: measures $μ$ and sets $Γ$ of lacunary form, admitting lacunary Fourier series in $L^2(μ)$; i.e., configurations $Γ$ arranged with a geometric progression of empty spacing, or missing parts, gaps. Given $Γ$, we find corresponding generalized Szeg\" o kernels $G_Γ$, and we compare them to the classical Szeg\" o kernel for $\bd$. Rather than the more traditional approach of starting with $μ$, and then asking for possibilities for sets $Γ$, such that we get Fourier series representations, we turn the problem upside down; now starting instead with a countably infinite discrete subset $Γ$, and, within a new duality framework, we study the possibilities for choices of measures $μ$.
Notes from a course taught by Palle Jorgensen in the fall semester of 2009. The course covered central themes in functional analysis and operator theory, with an emphasis on topics of special relevance to such applications as representation theory, harmonic analysis, mathematical physics, and stochastic integration.
In this paper, we consider certain elements in von Neumann algebras generated by graph groupoids. In particular, we are interested in finitely supported elements, called graph operators. We study the characterizations for self-adjointness, the unitary property, hyponormality and normality of graph operators.
A resistance network is a connected graph $(G,c)$. The conductance function $c_{xy}$ weights the edges, which are then interpreted as conductors of possibly varying strengths. The Dirichlet energy form $\mathcal E$ produces a Hilbert space structure ${\mathcal H}_{\mathcal E}$ on the space of functions of finite energy. The relationship between the natural Dirichlet form $\mathcal E$ and the discrete Laplace operator $Δ$ on a finite network is given by $\mathcal E(u,v) = \la u, \Lap v\ra_2$, where the latter is the usual $\ell^2$ inner product. We describe a reproducing kernel $\{v_x\}$ for $\mathcal E$ and used it to extends the discrete Gauss-Green identity to infinite networks: \[{\mathcal E}(u,v) = \sum_{G} u Δv + \sum_{\operatorname{bd}G} u \tfrac{\partial}{\partial \mathbf{n}} v,\] where the latter sum is understood in a limiting sense, analogous to a Riemann sum. This formula immediately yields a boundary sum representation for the harmonic functions of finite energy. Techniques from stochastic integration allow one to make the boundary $\operatorname{bd}G$ precise as a measure space, and give a boundary integral representation (in a sense analogous to that of Poisson or Martin boundary theory). This is done in terms of a Gel'fand triple $S \ci {\mathcal H}_{\mathcal E} \ci S'$ and gives a probability measure $\mathbb{P}$ and an isometric embedding of ${\mathcal H}_{\mathcal E}$ into $L^2(S',\mathbb{P})$, and yields a concrete representation of the boundary as a set of linear functionals on $S$.
We consider a family of measures $μ$ supported in $\br^d$ and generated in the sense of Hutchinson by a finite family of affine transformations. It is known that interesting sub-families of these measures allow for an orthogonal basis in $L^2(μ)$ consisting of complex exponentials, i.e., a Fourier basis corresponding to a discrete subset $Γ$ in $\br^d$. Here we offer two computational devices for understanding the interplay between the possibilities for such sets $Γ$ (spectrum) and the measures $μ$ themselves. Our computations combine the following three tools: duality, discrete harmonic analysis, and dynamical systems based on representations of the Cuntz $C^*$-algebras $\mathcal O_N$.
In this paper, we consider the relation between Toeplitz operators and elements in von Neumann algebras generated by certain graph groupoids.
A resistance network is a connected graph $(G,c)$. The conductance function $c_{xy}$ weights the edges, which are then interpreted as conductors of possibly varying strengths. The Dirichlet energy form $\mathcal E$ produces a Hilbert space structure (which we call the energy space ${\mathcal H}_{\mathcal E}$) on the space of functions of finite energy. We use the reproducing kernel $\{v_x\}$ constructed in \cite{DGG} to analyze the effective resistance $R$, which is a natural metric for such a network. It is known that when $(G,c)$ supports nonconstant harmonic functions of finite energy, the effective resistance metric is not unique. The two most natural choices for $R(x,y)$ are the ``free resistance'' $R^F$, and the ``wired resistance'' $R^W$. We define $R^F$ and $R^W$ in terms of the functions $v_x$ (and certain projections of them). This provides a way to express $R^F$ and $R^W$ as norms of certain operators, and explain $R^F \neq R^W$ in terms of Neumann vs. Dirichlet boundary conditions. We show that the metric space $(G,R^F)$ embeds isometrically into ${\mathcal H}_{\mathcal E}$, and the metric space $(G,R^W)$ embeds isometrically into the closure of the space of finitely supported functions; a subspace of ${\mathcal H}_{\mathcal E}$. Typically, $R^F$ and $R^W$ are computed as limits of restrictions to finite subnetworks. A third formulation $R^{tr}$ is given in terms of the trace of the Dirichlet form $\mathcal E$ to finite subnetworks. A probabilistic approach shows that in the limit, $R^{tr}$ coincides with $R^F$. This suggests a comparison between the probabilistic interpretations of $R^F$ vs. $R^W$.
In this paper we outline several points of view on the interplay between discrete and continuous wavelet transforms; stressing both pure and applied aspects of both. We outline some new links between the two transform technologies based on the theory of representations of generators and relations. By this we mean a finite system of generators which are represented by operators in Hilbert space. We further outline how these representations yield sub-band filter banks for signal and image processing algorithms.