arXiv subjects

Masaya Kato

Publications and source records attributed to Masaya Kato.

9 recordsLinked to original sources

arXiv2026

Validation-Centric AI-Assisted GPU Porting of a 250,000+ Line Legacy Weather Simulation Code

Recent advances in large language models have made CLI-based AI agents a practical tool for accelerating GPU porting of large legacy scientific applications. Such applications, however, are not merely old code bases; they are scientific assets whose credibility has been accumulated through long-term development, comparison with observations, and use in domain studies. GPU porting must therefore preserve this scientific validity while adapting the implementation to GPU-centric HPC systems. This paper presents a validation-centric AI-assisted GPU porting workflow through a case study of CReSS, a legacy Fortran weather simulation code with more than 250,000 lines. The workflow uses an AI agent to extract OpenMP regions, generate dump-based kernel benchmarks from physically meaningful simulation states, apply OpenACC transformations, and validate results through element-wise comparison with dumped reference data and application-level validation. Using a real typhoon simulation, the workflow produced numerically validated GPU implementations for 162 target kernels and achieved a 5.1x application-level speedup within practical wall-clock development cost. In particular, it detected numerical discrepancies in five kernels caused by floating-point and intrinsic-function differences, including threshold-sensitive branch divergence and cancellation effects, enabling feedback to the application developers. The case study suggests that, for large legacy scientific applications requiring dump-based validation, practical AI-assisted GPU porting must manage session-spanning context, runtime-state reconstruction, and costly recovery from small static-analysis omissions. These findings demonstrate that AI-assisted GPU porting requires not only code generation, but validation-centric workflow design.

cs.DC↗

arXiv2026

Autofocus Method for Human-Body Imaging under Respiratory Motion Using Synthetic Aperture Radar

This study presents an effective autofocusing approach for synthetic aperture radar imaging of the human body under conditions of respiratory motion. The proposed method suppresses respiratory-motion-induced phase errors by separating radar echoes in the spatial- and time-frequency domains and estimating phase errors individually for each separated echo. By compensating for the estimated phase errors, synthetic aperture radar images focused on all scattering points are generated, even when multiple body parts exhibit different motions due to respiration. The performance of the proposed method is evaluated through experiments with four participants in the supine position. Compared with a conventional method, the proposed approach improves image quality by a factor of 5.1 in terms of Muller-Buffington sharpness, and reduces the root-mean-square error with respect to a reference point cloud from 34 mm to 20 mm.

eess.SP↗

arXiv2025

Self-localization on a 3D map by fusing global and local features from a monocular camera

Self-localization on a 3D map by using an inexpensive monocular camera is required to realize autonomous driving. Self-localization based on a camera often uses a convolutional neural network (CNN) that can extract local features that are calculated by nearby pixels. However, when dynamic obstacles, such as people, are present, CNN does not work well. This study proposes a new method combining CNN with Vision Transformer, which excels at extracting global features that show the relationship of patches on whole image. Experimental results showed that, compared to the state-of-the-art method (SOTA), the accuracy improvement rate in a CG dataset with dynamic obstacles is 1.5 times higher than that without dynamic obstacles. Moreover, the self-localization error of our method is 20.1% smaller than that of SOTA on public datasets. Additionally, our robot using our method can localize itself with 7.51cm error on average, which is more accurate than SOTA.

cs.RO↗

arXiv2024

Radar-based Measurement of the Body Movements of Multiple Students in Classroom Environments

We demonstrate the feasibility of the radar-based measurement of body movements in scenarios involving multiple students using a pair of 79-GHz millimeter-wave radar systems with array antennas. We quantify the body motion using the Doppler frequency calculated from radar echoes. The measurement accuracy is evaluated for two experimental scenarios, namely university students in an office and elementary school students in a classroom. The body movements measured using the two radar systems are compared to evaluate the repeatability and angle dependency of the measurement. Moreover, in the first scenario, we compare the radar-estimated body movement with subjective evaluation scores provided by two evaluators. In the first scenario, the coefficient of correlation between the radar-estimated body movement and the subjective evaluation score is 0.73 on average, with a maximum value of 0.97; in the second scenario, the average correlation coefficient of body movements measured using two radar systems is as high as 0.78. These results indicate that the proposed approach can be used to monitor the body movements of multiple students in realistic scenarios.

eess.SP↗

arXiv2017

Machine learning technique to find quantum many-body ground states of bosons on a lattice

We develop a variational method to obtain many-body ground states of the Bose-Hubbard model using feedforward artificial neural networks. A fully-connected network with a single hidden layer works better than a fully-connected network with multiple hidden layers, and a multi-layer convolutional network is more efficient than a fully-connected network. AdaGrad and Adam are optimization methods that work well. Moreover, we show that many-body ground states with different numbers of atoms can be generated by a single network.

cond-mat.dis-nn↗

arXiv2017

Moving obstacle potential in a spin-orbit-coupled Bose-Einstein condensate

We investigate the dynamics around an obstacle potential moving in the plane-wave state of a pseudospin-$1/2$ Bose-Einstein condensate with Rashba spin-orbit coupling. We numerically investigate the dynamics of the system and find that it depends not only on the velocity of the obstacle but also significantly on the direction of obstacle motion, which are verified by a Bogoliubov analysis. The excitation diagram with respect to the velocity and direction is obtained. The dependence of the critical velocity on the strength of the spin-orbit coupling and the size of the obstacle is also investigated.

cond-mat.quant-gas↗

arXiv2017

Vortex pairs in a spin-orbit coupled Bose-Einstein condensate

Static and dynamic properties of vortices in a two-component Bose-Einstein condensate with Rashba spin-orbit coupling are investigated. The mass current around a vortex core in the plane-wave phase is found to be deformed by the spin-orbit coupling, and this makes the dynamics of the vortex pairs quite different from those in a scalar Bose-Einstein condensate. The velocity of a vortex-antivortex pair is much smaller than that without spin-orbit coupling, and there exist stationary states. Two vortices with the same circulation move away from each other or unite to form a stationary state.

cond-mat.quant-gas↗

arXiv2016

Spin-orbit-coupled Bose-Einstein condensates held under toroidal trap

We study a quasispin-$1/2$ Bose-Einstein condensate with synthetically generated spin-orbit coupling in a toroidal trap, and show that the system has a rich variety of ground and metastable states. As the central hole region increases, i.e., the potential changes from harmonic-like to ring-like, the condensate exhibits a variety of structures, such as triangular stripes, flower-petal patterns, and counter-circling states. We also show that the rotating systems have exotic vortex configurations. In the limit of a quasi-one dimensional ring, the quantum many-body ground state is obtained, which is found to be the fragmented condensate.

cond-mat.quant-gas↗

arXiv2016

Twisted spin vortices in a spinor-dipolar Bose-Einstein condensate with Rashba spin-orbit coupling

We consider a spin-1 Bose-Einstein condensate with Rashba spin-orbit coupling and dipole-dipole interaction confined in a cigar-shaped trap. Due to the combined effects of spin-orbit coupling, dipole-dipole interaction, and trap geometry, the system exhibits a rich variety of ground-state spin structures, including twisted spin vortices. The ground-state phase diagram is determined with respect to the strengths of the spin-orbit coupling and dipole-dipole interaction.

cond-mat.quant-gas↗