RKHS Representation of Algebraic Convolutional Filters with Integral Operators
Authors: Alejandro Parada-Mayorga, Alejandro Ribeiro, Juan Bazerque
Organizations: Dept. of Electrical Eng., Univ.of Colorado (Denver). USA · Dept. of Electrical and Systems Eng., Univ.of Pennsylvania. · Dept. of Electrical and Comput Eng., Univ.of Pittsburgh. USA
Integral operators play a central role in signal processing, underpinning classical convolution, and filtering on continuous network models such as graphons. While these operators are traditionally analyzed through spectral decompositions, their connection to reproducing kernel Hilbert spaces (RKHS) has not been systematically explored within the algebraic signal processing framework. In this paper, we develop a comprehensive theory showing that the range of integral operators naturally induces RKHS convolutional signal models whose reproducing kernels are determined by a box product of the operator symbols. We characterize the algebraic and spectral properties of these induced RKHS and show that polynomial filtering with integral operators corresponds to iterated box products, giving rise to a unital kernel algebra. This perspective yields pointwise RKHS representations of filters via the reproducing property, providing an alternative to operator-based implementations. Our results establish precise connections between eigendecompositions and RKHS representations in graphon signal processing, extend naturally to directed graphons, and enable novel spatial--spectral localization results. Furthermore, we show that when the spectral domain is a subset of the original domain of the signals, optimal filters for regularized learning problems admit finite-dimensional RKHS representations, providing a principled foundation for learnable filters in integral-operator-based neural architectures.
Figures & tables
Figure 1: Illustration for Example 2 . Left: Discrete representation showing eigenfunctions φ1(u) , φ4(u) , and φ7(u) (curves) evaluated at sample points u∈{0.2,0.45,0.7,0.86} (markers), with coefficient vector α=[−2,1,−0.5,0.2] indicated by αv . Center: Graphon signal f(u) and kernel functions kv(u) for v∈{0.2,0.45,0.7,0.86} as functions of u∈[0,1] . Right: Normalized Fourier coefficients fi/λi2 .
Figure 2: Pictorial representation of a generic algebraic signal model (ASM) (A,M,ρ) . Filters are elements of the algebra A while the signals are the elements of the vector space H . The homomorphism, ρ , translates the abstract filters in A into concrete operators in End(H) , that act on the signals in H . The symbol End(H) represents the space of linear operators from H onto itself.
Figure 3: Illustration diagram of the results in Theorem 9 and Corollary 2 . The operator TKr maps the signal f∈H(K) on the RKHS H(K□(r+1)) , where the resultant diffused signal can be written as an expansion in terms of the kernel functions kℓr(r+1) . Then, the result of diffusing f with p(TK)=∑r=0RhrTKrf is obtained as a weighted sum of signals on the RKHS spaces {H(K□(r+1))}r=0R , where the weights are the filter coefficients {hr}r=0R .
Figure 4: Diagramatic illustration of the consequences of Corollary 4 . An RKHS with a finite expansion in terms of the kt functions cannot be exactly bandlimited. However, due to the nature of the eigenvalues, σi , of the operator TK the signal can be approximately bandlimited. Given the decomposition of a signal f∈H(K) with the representation in ( 52 ), the tradeoff between RKHS finiteness and bandlimited frequencies is determined by ∣T∣−B . If B<∣T∣ it is possible to choose ∣T∣−B coefficients at to minimize the size of fB+1:∣T∣ in ( 52 ). Then, B coefficients determine a specific low pass behavior, while the remaining coefficients are selected to reduce the size of the residuals (green shaded area on the right side). This implies that, the larger ∣T∣−B is, the faster the residuals can go to zero.
Figure 5: Illustration associated to Example 3 . Using the representer theorem we find the optimal polynomial filter p(t) , such that p(σi)=exp(−(σi−σc)2/γ) where σc and γ are fixed parameters. From Corollary 5 , the optimal filter can be written as p∗(u)=∑i=1qaikσi(u) . The pictures show the resultant filter, when considering multiple values of q – the number of points used in the regression – and multiple values of σc (the localization of the maximum point of the filters).
Reproducing kernel Hilbert spaces (RKHSs) use kernels to represent and regularize functions, enabling tractable learning, inference, and approximation.
Graph signal processing tasks that leverage spectral information typically assume access to the complete graph topology, which is often unavailable in practice. We propose a systematic framework for subgraph filter learning (SFL), where subgraph-supported operators approximate ambient graph filters under partial observations. We formulate SFL as a statistical learning problem in which optimal subgraph operators are inherently data-dependent. To address the difficulty of directly estimating such operators, we develop a subgraph filter algebra based on distance-aware Laplacian constructions, defining a structured and controllable class of filters for effective approximation. We further establish performance risk bounds under the least squares loss, quantifying how well the learned operator approximates the restricted ambient mapping. Experiments real-world datasets show that, for SFL tasks, the proposed algebraic models consistently outperform polynomial filters, distribution-agnostic operators, and direct numerical filter learning baselines that attempt to recover the underlying structure from data.
We develop a stochastic approximation framework for learning nonlinear operators between infinite-dimensional spaces utilizing general Mercer operator-valued kernels. Our framework encompasses two key classes: (i) operator-valued kernels whose associated integral operators are compact and hence admit discrete spectral decompositions, and (ii) separable kernels of the form K(x,x′)=k(x,x′)T, where k is a scalar-valued kernel and T is a positive operator on the output space. This broad setting induces expressive vector-valued reproducing kernel Hilbert spaces (RKHSs) that generalize the classical K=kI paradigm, thereby enabling rich structural modeling with rigorous theoretical guarantees. To address target operators lying outside the RKHS, we introduce vector-valued interpolation spaces to precisely quantify misspecification error. Within this framework, we establish non-asymptotic convergence rates for prediction, estimation, and misspecification errors in the online and finite-horizon settings. Importantly, the framework also accommodates a range of operator learning settings, from Fredholm integral operators to encoder--decoder architectures. Numerical experiments on the two-dimensional Navier--Stokes equations illustrate the proposed approach.
Jia-Qi Yang, Lei Shi
School of Mathematical Sciences and Shanghai Key Laboratory for Contemporary Applied Mathematics, Fudan University, Shanghai 200433, China.
The convolution operator is the fundamental building block of modern convolutional neural networks (CNNs), owing to its simplicity, translational equivariance, and efficient implementation. However, its structure as a fixed, linear, locally-averaging operator limits its ability to capture structured signal properties such as low-rank decompositions, adaptive basis representations, and non-uniform spatial dependencies. This paper presents a systematic taxonomy of operators that extend or replace the standard convolution in learning-based image processing pipelines. We organise the landscape of alternative operators into five families: (i) decomposition-based operators, which separate structural and noise components through singular value or tensor decompositions; (ii) adaptive weighted operators, which modulate kernel contributions as a function of spatial position or signal content; (iii) basis-adaptive operators, which optimise the analysis bases together with the network weights; (iv) integral and kernel operators, which generalise the convolution to position-dependent and non-linear kernels; and (v) attention-based operators, which relax the locality assumption entirely. For each family, we provide a formal definition, a discussion of its structural properties with respect to the convolution, and a critical analysis of the tasks for which the operator is most appropriate. We further provide a comparative analysis of all families across relevant dimensions -- linearity, locality, equivariance, computational cost, and suitability for image-to-image and image-to-label tasks -- and outline the open challenges and future directions of this research area.