arXiv subjects

Masahiro Ikeda

Publications and source records attributed to Masahiro Ikeda.

At least 19 recordsLinked to original sources

arXiv2026

Cahn-Hilliard equation associated with hypergraph

A hypergraph Laplacian was introduced to investigate the structure of networks written as hypergraphs. It was shown that the behavior of solutions to an evolution equation associated with the hypergraph Laplacian quite resembles that of solutions to the classical heat equation. Hence by replacing the Laplacian in PDEs with the hypergraph Laplacian, we might be able to introduce various diffusion models on hypergraphs (discrete domains), whose behavior of solution is similar to PDEs. In this paper, we consider a system of equations obtained by replacing the Laplacian in the original Cahn--Hilliard equation with the hypergraph Laplacian. Due to the nonlinearity and multivaluedness of the hypergraph Laplacian, it is difficult to apply methods for the original Cahn--Hilliard equation to our problem. To cope with these difficulty, we shall introduce a new proof by using properties of the hypergraph Laplacian in this paper.

math.CA↗

arXiv2026

On solitary wave solutions with two-frequency parameters to the three-component system of quadratic nonlinear Schrödinger equations

In the present paper, we consider the Cauchy problem of a system of three nonlinear Schrödinger equations with quadratic nonlinearity. We first prove the existence of ground states in the form of solitary wave solutions with two frequency parameters. We then show that the conditions on the frequency parameters for the existence of ground states depend on the resonance structure of the system. Next, we give the two results for global solutions. The first is an improvement of global well-posedness for initial data below the ground state threshold. The second is an improvement of global well-posedness for oscillating initial data. We also prove the orbital stability of the ground state sets with small speed parameter.

math.AP↗

arXiv2026

Global existence of solutions for nonlinear damped wave equations in an exterior domain with nonlinearities of derivative type

In this paper, we are interested in considering the semi-linear damped wave equation with the power nonlinearity of derivative type in $2$D exterior domains to indicate the global (in time) existence of small data solutions, which has never appeared in previous studies. Our proof is based on constructing a judicious time-weighted function space and demonstrating nonlinear estimates compatible with fractional powers of the Dirichlet Laplacian linked to the semigroup decay structure. Moreover, the large time behavior of the time derivative of the obtained global solutions and its sharp estimate are also discussed in this paper.

math.AP↗

arXiv2026

Boundedness of composition operator in Orlicz-Morrey spaces

In this paper, we investigate necessary and sufficient conditions on the boundedness of composition operators on the Orlicz-Morrey spaces. The results of boundedness include Lebesgue and generalized Morrey spaces as special cases. Further, we characterize the boundedness of composition operators on the weak Orlicz-Morrey spaces. The weak Orlicz-Morrey spaces contain the Orlicz-Morrey spaces.

math.FA↗

arXiv2026

Ghosts in Neural Networks: Existence, Structure and Role of Infinite-Dimensional Null Space

We study parameter nonuniqueness in continuous-width depth-two fully connected neural networks. Our main contribution is a direct method for solving the neural-network equation $S[γ]=f$. Starting from the Fourier expression of the synthesis operator, separation of variables produces a ridgelet particular solution and identifies every homogeneous direction. To isolate the argument, we first prove an abstract reconstruction formula for unitary factorizations, yielding the adjoint, normalized right inverse, and orthogonal solution geometry. We then specialize this formula to neural-network synthesis: for tempered-distribution activations such as ReLU, we equip the activation class $A_{s,t}$ with a Hilbert structure, construct compatible coefficient and parameter Hilbert spaces $H_{s,t}$ and $G_{s,t}$, and prove that $S:G_{s,t}\to L^2(\mathbb R^m)$ is bounded. The resulting ridgelet expansion exhausts the null space and the complete solution set and identifies the unique minimum-norm parameter distribution. Concrete examples give adjoint ridgelet functions for standard activations. Further developments show that finite-measure null elements admit normalized width-$N$ discretizations with $O(N^{-1/2})$ output error and characterize how additive parameter perturbations can reveal information encoded in the null space. A Lean 4 blueprint for the main results is available at https://shosonoda.github.io/lean-ridgelet/ .

cs.LG↗

arXiv2026

Unconditional uniqueness of Hardy--Hénon parabolic equations on Herz spaces

In this paper, we introduce the unconditional uniqueness of solutions in Herz spaces for the Hardy--Hénon parabolic equation, which is a semilinear heat equation with a power-type weight in the nonlinear term $|x|^γ|u|^{α-1}u$. It is expected that the power-type weight in the nonlinear term can be effectively handled within Herz spaces. In fact, our result in Herz spaces $\dot{K}^s_{q,r}({\mathbb R}^n)$ relaxes the endpoint case $q=α$ and the large interpolation exponent case $r\ge q$ compared to previous results.

math.AP↗

arXiv2026

Spectral Truncation Kernels: Noncommutativity in $C^*$-algebraic Kernel Machines

A central question in vector- and function-valued learning is how to design kernels that capture both local and non-local interactions while remaining computationally tractable. Existing operator-valued kernels offer only partial answers: separable kernels are efficient but fail to model interactions across the function domain, while commutative kernels capture only pointwise structure. To address this, we propose spectral truncation kernels, a new class of positive definite kernels for vector- and function-valued learning based on spectral truncation and $C^*$-algebra. By allowing noncommutative products in the kernel construction, the proposed kernels induce interactions across the data function domain and fill the gap between existing separable and commutative kernels. In addition, by using the $C^*$-algebraic framework, we reduce the computational cost compared to the existing vector-valued RKHS framework with operator-valued kernels.

stat.ML↗

arXiv2026

Ground states of the defocusing nonlinear Schrödinger equation with a point interaction in dimensions 2 and 3

This paper is concerned with ground states of the defocusing nonlinear Schrödinger equation with a point interaction, \[ \mathrm{i} \partial_t ψ= -Δ_αψ+ ψ|ψ|^{p - 2} \quad \text{in} \quad \mathbb{R} \times \mathbb{R}^N, \] where $- Δ_α$ denotes the Laplacian of point interaction centered at the origin with inverse s-wave scattering length $- 2 (N - 1) πα$ and we suppose that either (i) $N = 2$, $α\in \mathbb{R}$ and $p > 2$ or (ii) $N = 3$, $α< 0$ and $2 < p < 3$. At sufficiently small masses, (i) we prove that this equation admits ground states, (ii) we obtain some qualitative properties of ground states and (iii) we obtain some results relating ground states with critical points of the associated action functional.

math.AP↗

arXiv2026

Why and When Deep is Better than Shallow: Implementation-Agnostic State-Transition Model of Deep Learning

Why and when does depth improve generalization? We study this question in an implementation-agnostic state-transition model, where a depth-$k$ predictor is a readout class $H$ composed with the word ball $B(k,F)$ generated by hidden state transitions. Generalization bounds separate implementation error, approximation error, and statistical complexity, and upper bound the depth-dependent variance term by a Dudley entropy integral over $B(k,F)$, with a conditional lower-bound diagnostic under readout separation. We identify geometric and semigroup mechanisms that keep this entropy contribution saturated or polynomial, and contrast them with separation mechanisms that recover the classical exponential-growth obstruction. Coupling these variance upper bounds with approximation rates gives typical depth trade-off patterns, clarifying that depth is statistically favorable when approximation improves rapidly while the transition semigroup remains geometrically tame.

cs.LG↗

arXiv2026

On the Hardy-Hénon heat equation with an inverse square potential

We study Cauchy problem for the Hardy-Hénon parabolic equation with an inverse square potential, namely, \[\partial_tu -Δu+a|x|^{-2} u= |x|^γ F_α(u),\] where $a\ge-(\frac{d-2}{2})^2,$ $γ\in \mathbb R$, $α>1$ and $F_α(u)=μ|u|^{α-1}u, μ|u|^α$ or $μu^α$, $μ\in \{-1,0,1\}$. We establish sharp fixed time-time decay estimates for heat semigroups $e^{-t (-Δ+ a|x|^{-2})}$ in weighted Lebesgue spaces. This may be of independent interest. As an application, we establish local well-posedness in scale subcritical and critical weighted Lebesgue spaces and small data global existence in critical weighted Lebesgue spaces. Further, under certain conditions on $γ$ and $α,$ we show that local solution cannot be extended to global one for certain initial data in the subcritical regime. Thus, finite time blow-up in the subcritical Lebesgue space norm is exhibited. We also demonstrate nonexistence of local positive weak solution (and hence failure of local well-posedness) in supercritical case for $α>1+\frac{2+γ}{d}$ the Fujita exponent.

math.AP↗

arXiv2026

Why High-rank Neural Networks Generalize?: An Algebraic Framework with RKHSs

We derive a new Rademacher complexity bound for deep neural networks using Koopman operators, group representations, and reproducing kernel Hilbert spaces (RKHSs). The proposed bound describes why the models with high-rank weight matrices generalize well. Although there are existing bounds that attempt to describe this phenomenon, these existing bounds can be applied to limited types of models. We introduce an algebraic representation of neural networks and a kernel function to construct an RKHS to derive a bound for a wider range of realistic models. This work paves the way for the Koopman-based theory for Rademacher complexity bounds to be valid for more practical situations.

cs.LG↗

arXiv2025

Koopman operators with intrinsic observables in rigged reproducing kernel Hilbert spaces

This paper presents a novel approach for estimating the Koopman operator defined on a reproducing kernel Hilbert space (RKHS) and its spectra. We propose an estimation method, what we call Jet Extended Dynamic Mode Decomposition (JetEDMD), leveraging the intrinsic structure of RKHS and the geometric notion known as jets to enhance the estimation of the Koopman operator. This method refines the traditional Extended Dynamic Mode Decomposition (EDMD) in accuracy, especially in the numerical estimation of eigenvalues. This paper proves JetEDMD's superiority through explicit error bounds and convergence rate for special positive definite kernels, offering a solid theoretical foundation for its performance. We also investigate the spectral analysis of the Koopman operator, proposing the notion of an extended Koopman operator within a framework of a rigged Hilbert space. This notion leads to a deeper understanding of estimated Koopman eigenfunctions and capturing them outside the original function space. Through the theory of rigged Hilbert space, our study provides a principled methodology to analyze the estimated spectrum and eigenfunctions of Koopman operators, and enables eigendecomposition within a rigged RKHS. We also propose a new effective method for reconstructing the dynamical system from temporally-sampled trajectory data of the dynamical system with solid theoretical guarantee. We conduct several numerical simulations using the van der Pol oscillator, the Duffing oscillator, the Hénon map, and the Lorenz attractor, and illustrate the performance of JetEDMD with clear numerical computations of eigenvalues and accurate predictions of the dynamical systems.

math.DS↗

arXiv2025

Generalized Stochastic Resilience for Early Warning Signals Based on Koopman Operator

Developing methods for detecting tipping phenomena at an early stage is an important problem in various fields such as ecology, medicine, and economics. A tipping phenomenon is characterized by a rapid transition resulting from the accumulation of small parameter changes and is known to be related to bifurcations of dynamical systems. However, few studies have examined how nonlinear properties near bifurcation points affect early warning signal (EWS) performance. In this study, we apply the Koopman operator, which describes the time evolution of dynamical systems in an infinite-dimensional function space, to generalize stochastic resilience the theoretical basis of EWSs such as variance-based ones. As a result, we develop a novel signal capable of more accurately predicting tipping events by separately isolating stochastic fluctuations induced by noise and contributions from a continuous spectrum emerging immediately above tipping points. Our experimental results demonstrate that the proposed approach detects early signs of tipping phenomena more robustly than conventional methods.

math.DS↗

arXiv2025

Asymptotically self-similar global solutions for Hardy-Hénon parabolic equations

We construct asymptotically self-similar global solutions to the Hardy-Hénon parabolic equation $\partial_t u - Δu = \pm |x|^γ |u|^{α-1} u$, $α>1$, $γ\in \mathbb{R}$ for a large class of initial data belonging to weighted Lorentz spaces. The solution may be asymptotic to a self-similar solution of the linear heat equation or to a self-similar solution to the Hardy-Hénon parabolic equation depending on the speed of decay of the initial data at infinity. The asymptotic results are new for the Hénon case $γ>0$. We also prove the stability of the asymptotic profiles. Our approach applies for $γ> -\min(2,d)$ and unifies the cases $γ>0$, $γ=0$ and $-\min(2,d)<γ<0$. For complex-valued initial data, a more intricate asymptotic behaviors can be shown; if either one of the real part or the imaginary part of the initial data has a faster spatial decay, then the solution exhibits a combined Nonlinear-"Modified Linear" asymptotic behavior, which is completely new even for the Fujita case $γ=0$. In Appendix, we show the non-existence of local positive solutions for supercritical initial data.

math.AP↗

arXiv2025

Critical curve for the weakly coupled system of damped wave equations with mixed nonlinearities

In this paper, we would like to consider the Cauchy problem for a weakly coupled system of semi-linear damped wave equations with mixed nonlinear terms. Our main objective is to draw conclusions about the critical curve of this problem using tools from Harmonic Analysis. Precisely, we obtain a new critical curve $pq = 1+ \frac{2}{n}$ for $n =1,2$ by proving global (in time) existence of small data Sobolev solutions when $pq > 1 +\frac{2}{n}$ and blow-up of weak solutions in finite time even for small data when $pq < 1+ \frac{2}{n}$ for $n \geq 1$. From this, we infer the impact of the nonlinearities of time derivative-type on the critical curve associated with the system.

math.AP↗

arXiv2025

Stability of standing waves for all frequencies to nonlinear Schrödinger equations with potentials in one dimension

In this paper, we study the orbital stability of standing waves for one-dimensional nonlinear Schrödinger equations with potentials. We show that the standing waves are orbitally stable for all frequencies in the $L^{2}$- subcritical and critical cases. Since the presence of potentials breaks the scale invariance of the equations, it is a delicate problem to apply the abstract theory of Grillakis, Shatah, and Strauss (1987) directly without a perturbative argument. For this reason, little is known about the orbital stability of standing waves for \textit{all} frequencies in the non-scale-invariant setting. We overcome this difficulty by employing the approach of Noris, Tavares, and Verzini (2014).

math.AP↗

arXiv2025

Dynamics of dissipative solutions to the Hardy-Sobolev parabolic equation

We study the long-time behaviour of solutions to the Hardy-Sobolev parabolic equation in critical function spaces for any spatial dimension $d \geq 5$. By employing the Fourier splitting method, we establish precise decay rates for dissipative solutions, meaning those whose critical norm vanishes as time approaches infinity. Our findings offer a deeper understanding of the asymptotic properties and dissipation mechanisms governing this equation.

math.AP↗

arXiv2025

Deep Ridgelet Transform and Unified Universality Theorem for Deep and Shallow Joint-Group-Equivariant Machines

We present a constructive universal approximation theorem for learning machines equipped with joint-group-equivariant feature maps, called the joint-equivariant machines, based on the group representation theory. ``Constructive'' here indicates that the distribution of parameters is given in a closed-form expression known as the ridgelet transform. Joint-group-equivariance encompasses a broad class of feature maps that generalize classical group-equivariance. Particularly, fully-connected networks are not group-equivariant but are joint-group-equivariant. Our main theorem also unifies the universal approximation theorems for both shallow and deep networks. Until this study, the universality of deep networks has been shown in a different manner from the universality of shallow networks, but our results discuss them on common ground. Now we can understand the approximation schemes of various learning machines in a unified manner. As applications, we show the constructive universal approximation properties of four examples: depth-$n$ joint-equivariant machine, depth-$n$ fully-connected network, depth-$n$ group-convolutional network, and a new depth-$2$ network with quadratic forms whose universality has not been known.

cs.LG↗