arXiv subjects

Toru Takahashi

Publications and source records attributed to Toru Takahashi.

14 recordsLinked to original sources

arXiv2026

Toward AI That Understands Self and Others: A World-Model Theory of Cognitive Diversity and Alignment

Modern societies possess more information than ever before, yet they do not converge toward a single shared understanding. The same events, facts, laws, technologies, or risks can be interpreted as evidence of freedom, danger, exclusion, injustice, responsibility, or unrealized possibility. Existing discussions often treat such disagreement as a conflict of values, preferences, or beliefs. This paper argues that disagreement is already a late-stage phenomenon. The central premise is simple but not trivial: observation is not yet inference. Not every observation becomes inferentially relevant, and not every possible object in an observation sequence becomes an estimation target. A possible target becomes admissible only when a state representation can be constructed that is approximately sufficient for prediction, evaluation, or action with respect to that target. This paper develops a world-model theory of cognitive diversity and alignment by reconstructing recognition as the construction of such approximate sufficient statistics under finite informational, representational, observational, and action constraints. It formulates this position as the Multi-Phase Inference Assumption (MIA) and defines its core internal mechanism as the Multi-Phase Inference Mechanism (MIM). The framework introduces alignment maps and transformation loss to analyze how heterogeneous world models communicate without being collapsed into a single representation. World-model alignment is therefore processability, not agreement: the design of AI systems that help heterogeneous forms of intelligence remain mutually processable while preserving their distinct error-detection capacities.

cs.AI↗

arXiv2026

Why Conclusions Diverge from the Same Observations: Formalizing World-Model Non-Identifiability via an Inference

When people share the same documents and observations yet reach different conclusions, the disagreement often shifts into a judgment that the other party is cognitively defective, irrational, or acting in bad faith. This paper argues that such divergence is better described as a form of non-identifiability inherent in inference and learning, rather than as a defect of the other party. We organize the phenomenon into two levels: (i) $θ$-level non-identifiability, where conclusions diverge under the same world model $W$ because inference settings differ; and (ii) $W$-level non-identifiability, where repeated use of an inference setting $θ$ biases data exposure and update rules, causing the learned world model $W$ itself to diverge. We introduce an inference profile $θ= (R, E, S, D)$, consisting of Reference, Exploration, Stabilization, and Horizon, and show how outputs can split even for the same observation $o$ and the same $W$. We further explain why disagreements tend to project onto a small number of bases -- abstract versus concrete, externalizability, and order versus freedom -- as a consequence of general constraints on learning systems: computational, observational, and coordination constraints. Finally, we relate the framework to deep representation learning, including representation hierarchy, latent-state estimation, and regularization-exploration trade-offs, and illustrate the framework through a case study on AI regulation debates.

cs.AI↗

arXiv2025

A shape optimisation of mutual inductances among coils

This paper introduces a shape optimisation framework for achieving desired mutual inductances (MIs) among coils in 3D space. Utilising a wire modelling approach, the coils are discretised using B-spline curves, with control points (CPs) serving as design variables. The key contribution is the derivation of the shape derivative of the objective function in terms of MIs, enabling the use of gradient-based quasi-Newton optimisation methods. A coil length constraint is also incorporated. The study demonstrates the effectiveness of the framework through numerical examples, validating the theoretical and numerical developments. This approach addresses the largely unexplored area of magnetostatic MI optimisation within the wire modelling framework, offering a computationally efficient alternative to finite element methods etc.

math-ph↗

arXiv2024

Designing gradient coils with the shape derivative and the closed B-spline curves

This study proposes a versatile and efficient optimisation method for discrete coils that induce a magnetic field by their steady currents. The prime target is gradient coils for MRI (Magnetic Resonance Imaging). The derivative (gradient) of the $z$-component the magnetic field, which is calculated by the Biot--Savart's law, with respect to the $z$-coordinate in the Cartesian $xyz$ coordinate system is considered as the objective function. Then, the derivative of the objective function with respect to a change of coils in shape is formulated according to the concept of shape optimisation. The resulting shape derivative (as well as the Biot--Savart's law) is smoothly discretised with the closed B-spline curves. In this case, the control points (CPs) of the curves are naturally selected as the design variables. As a consequence, the shape derivative is discretised to the sensitivities of the objective function with respect to the CPs. Those sensitivities are available to solve the present shape-optimisation problem with a certain gradient-based nonlinear-programming solver. The numerical examples exhibit the mathematical reliability, computational efficiency, and engineering applicability of the proposed methodology based on the shape derivative/sensitivities and the closed B-spline curves.

physics.med-ph↗

arXiv2023

A fast time-domain boundary element method for three-dimensional electromagnetic scattering problems

This paper proposes a fast time-domain boundary element method (TDBEM) to solve three-dimensional transient electromagnetic scattering problems regarding perfectly electric conductors in the classical marching-on-in-time manner. The algorithm of the fast TDBEM is a time-domain variant of the interpolation-based fast multipole method (IFMM), which is similar to the time-domain IFMM for acoustic scattering problems investigated in the author's previous studies. The principle of the present IFMM is to interpolate the kernel functions of the electric and magnetic field integral equations (EFIE and MFIE, respectively) so that every kernel function is expressed in a form of separation of variables in terms of both the spatial and temporal variables. Such an expression enables to construct a fast method to evaluate the scalar and vector potentials in the EFIE and MFIE with using so-called multipole-moments and local-coefficients associated with a space-time hierarchy. As opposed to $O(N_s^2 N_t)$ of the conventional TDBEM, the computational complexity of the fast TDBEM is estimated as $O(N_s^{1+δ}N_t)$, where $N_s$ and $N_t$ stand for the spatial and temporal degrees of freedom, respectively, and $δ$ is typically $1/2$ or $1/3$. The numerical examples presented the advantages of the proposed fast TDBEM over the conventional TDBEM when solving large-scale problems.

physics.comp-ph↗

arXiv2022

A topology optimization of open acoustic waveguides based on a scattering matrix method

This study presents a topology optimization scheme for realizing a bound state in the continuum along an open acoustic waveguide comprising a periodic array of elastic materials. First, we formulate the periodic problem as a system of linear algebraic equations using a scattering matrix associated with a single unit structure of the waveguide. The scattering matrix is numerically constructed using the boundary element method. Subsequently, we employ the Sakurai--Sugiura method to determine resonant frequencies and the Floquet wavenumbers by solving a nonlinear eigenvalue problem for the linear system. We design the shape and topology of the unit elastic material such that the periodic structure has a real resonant wavenumber at a given frequency by minimizing the imaginary part of the resonant wavenumber. The proposed topology optimization scheme is based on a level-set method with a novel topological derivative. We demonstrate a numerical example of the proposed topology optimization and show that it realizes a bound state in the continuum through some numerical experiments.

math.NA↗

arXiv2021

An enhancement of the fast time-domain boundary element method for the three-dimensional wave equation

Our objective is to stabilise and accelerate the time-domain boundary element method (TDBEM) for the three-dimensional wave equation. To overcome the potential time instability, we considered using the Burton--Miller-type boundary integral equation (BMBIE) instead of the ordinary boundary integral equation (OBIE), which consists of the single- and double-layer potentials. In addition, we introduced a smooth temporal basis, i.e. the B-spline temporal basis of order $d$, whereas $d=1$ was used together with the OBIE in a previous study [Takahashi 2014]. Corresponding to these new techniques, we generalised the interpolation-based fast multipole method that was developed in \cite{takahashi2014}. In particular, we constructed the multipole-to-local formula (M2L) so that even for $d\ge 2$ we can maintain the computational complexity of the entire algorithm, i.e. $O(N_{\rm s}^{1+δ} N_{\rm t})$, where $N_{\rm s}$ and $N_{\rm t}$ denote the number of boundary elements and the number of time steps, respectively, and $δ$ is theoretically estimated as $1/3$ or $1/2$. The numerical examples indicated that the BMBIE is indispensable for solving the homogeneous Dirichlet problem, but the order $d$ cannot exceed 1 owing to the doubtful cancellation of significant digits when calculating the corresponding layer potentials. In regard to the homogeneous Neumann problem, the previous TDBEM based on the OBIE with $d=1$ can be unstable, whereas it was found that the BMBIE with $d=2$ can be stable and accurate. The present study will enhance the usefulness of the TDBEM for 3D scalar wave problems.

math.NA↗

arXiv2021

An isogeometric boundary element method for three-dimensional doubly-periodic layered structures in electromagnetics

This paper proposes an isogeometric boundary element method (IGBEM) to solve the electromagnetic scattering problems for three-dimensional doubly-periodic multi-layered structures. The main concerns are the constructions of (i) an open surface (between two layers) and (ii) a vector basis function with using the B-spline functions. Regarding (i), we considered an algorithm to generate a doubly-periodic open surface with the tensor product of the B-spline functions of any degree. Regarding (ii), we employed the vector basis function based on the B-spline functions, which was proposed by Buffa et al. (2010), and adapted it to the underlying periodic problems so that it can satisfy the quasi-periodic condition on the boundary of an open surface. The proposed IGBEM worked for solving some numerical examples satisfactorily and proved the applicability to plasmonic simulations.

math.NA↗

arXiv2021

A shape optimisation with the isogeometric boundary element method and adjoint variable method for the three-dimensional Helmholtz equation

This paper presents a shape optimisation system to design the shape of an acoustically-hard object in the three-dimensional open space. Boundary element method (BEM) is suitable to analyse such an exterior field. However, the conventional BEM, which is based on piecewise polynomial shape and interpolation functions, can require many design variables because they are usually chosen as a part of the nodes of the underlying boundary element mesh. In addition, it is not easy for the conventional method to compute the gradient of the sound pressure on the surface, which is necessary to compute the shape derivative of our interest, of a given object. To overcome these issues, we employ the isogeometric boundary element method (IGBEM), which was developed in our previous work. With using the IGBEM, we can design the shape of surfaces through control points of the NURBS surfaces of the target object. We integrate the IGBEM with the nonlinear programming software through the adjoint variable method (AVM), where the resulting adjoint boundary value problem can be also solved by the IGBEM with a slight modification. The numerical verification and demonstration validate our shape optimisation framework.

math.NA↗

arXiv2019

Parallelization of the inverse fast multipole method with an application to boundary element method

We present an algorithm to parallelize the inverse fast multipole method (IFMM), which is an approximate direct solver for dense linear systems. The parallel scheme is based on a greedy coloring algorithm, where two nodes in the hierarchy with the same color are separated by at least $σ$ nodes. We proved that when $σ\ge 6$, the workload associated with one color is embarrassingly parallel. However, the number of nodes in a group (color) may be small when $σ= 6$. Therefore, we also explored $σ= 3$, where a small fraction of the algorithm needs to be serialized, and the overall parallel efficiency was improved. We implemented the parallel IFMM using OpenMP for shared-memory machines. Successively, we applied it to a fast-multipole accelerated boundary element method (FMBEM) as a preconditioner, and compared its efficiency with (a) the original IFMM parallelized by linking a multi-threaded linear algebra library and (b) the commonly used parallel block-diagonal preconditioner. Our results showed that our parallel IFMM achieved at most $4\times$ and $11\times$ speedups over the reference method (a) and (b), respectively, in realistic examples involving more than one million variables.

physics.comp-ph↗

arXiv2016

A fast topology optimisation for material- and geometry-independent cloaking devices with the BEM and the ${\mathcal H}$-matrix method

We show a design method of cloaking devices which work for target objects with arbitrary shape and material by a topology optimisation with an accurate and efficient sensitivity analysis. Most of past researches on topology optimisation of cloaking devices intend to hide a circle-shaped perfect electric conductor. In this case, the cloaking effect is highly dependent on the shape and material of a target object. In this study, we consider to design a cloaking device which work regardless of the property of target objects by modifying the definition of the objective function. Also, we developed an efficient and accurate sensitivity analysis with the boundary element method and the ${\mathcal H}$-matrix method. We show that the proposed method can successfully obtain desired cloaking devices with low computational cost.

math.NA↗

arXiv2016

ASTE Observation of Jupiter's Stratospheric Composition: Detection of Carbon Monosulfide ($J$=7--6) in 19 Years After the Cometary Impact

In Jupiter's stratosphere, gaseous carbon monosulfide (CS) was first discovered in 1994 by millimeter and ultraviolet observations as a product induced by the collision of comet Shoemaker-Levy 9 (SL9). {To constrain sulfur chemistry, in 2013, 19 years after the SL9 event, we observed Jupiter's stratospheric CS $J$=7 -- 6 rotational transition at 0.8 mm wavelength by using the Atacama Submillimeter Telescope Experiment (ASTE) 10-m single dish telescope. The CS molecular line was successfully detected with 120 mK intensity in the antenna temperature scale. The obtained CS total mass shows $\sim$90$\%$ decrease relative to that observed in 1998. From the line shape analysis, CS is suggested to be present above the 0.2$^{+0.4}_{-0.15}$ mbar pressure level, which is comparable to that of determined in 1998 which was provided by SL9 along with CS, was observed up to the 2.0 mbar pressure region. The discrepancy of CS vertical distribution with \ce{H2O} and the decrease in its total mass may be attributed to the chemical loss of CS in Jupiter's stratosphere lower than the 0.2 mbar altitude region.

astro-ph.EP↗

arXiv2016

A level-set-based topology optimisation for acoustic-elastic coupled problems with a fast BEM-FEM solver

This paper presents a structural optimisation method in three-dimensional acoustic-elastic coupled problems. The proposed optimisation method finds an optimal allocation of elastic materials which reduces the sound level on some fixed observation points. In the process of the optimisation, configuration of the elastic materials is expressed with a level set function, and the distribution of the level set function is iteratively updated with the help of the topological derivative. The topological derivative is associated with state and adjoint variables which are the solutions of the acoustic-elastic coupled problems. In this paper, the acoustic-elastic coupled problems are solved by a BEM-FEM coupled solver, in which the fast multipole method (FMM) and a multi-frontal solver for sparse matrices are efficiently combined. Along with the detailed formulations for the topological derivative and the BEM-FEM coupled solver, we present some numerical examples of optimal designs of elastic sound scatterer to manipulate sound waves, from which we confirm the effectiveness of the present method.

math.NA↗

arXiv2007

Effect of Relic Neutrino on Neutrino Pair Emission from Metastable Atoms

A possiblity of measuring the cosmic neutrino temperature $\sim 1.9 K$ and other important quantities such as the chemical potential $μ$ and the decoupling temperature $T_d$ is discussed, using the recently proposed process of photon irradiated neutrino pair emission from metastable atoms. The Pauli blocking effect of relic neutrinos reduces the rate by a large factor $\approx (1 + m_1/T_d)/4$ at the threshold of the lightest neutrino pair (of mass $2m_1$). Correction of linear order in $μ$ near the mass thresholds can be used to improve the constraint on the lepton asymmetry.

hep-ph↗