arXiv subjects

Hiroyuki Abe

Publications and source records attributed to Hiroyuki Abe.

At least 19 recordsLinked to original sources

arXiv2026

MSSM flavors from 7-brane configurations of magnetized SYM on $R^{1,3} \times (T^2)^3/(Z_2 \times Z'_2)$

We show that chiral matters in the minimal supersymmetric standard model (MSSM) with semi-realistic flavor structures can be obtained from 7-brane configurations of magnetized super Yang-Mills (SYM) theory on a toroidal orbifold $R^{1,3} \times (T^2)^3/(Z_2 \times Z'_2)$, where background magnetic fluxes and Wilson-lines are turned on preserving four-dimensional ${\cal N}=1$ supersymmetry. The zero-mode spectrum of chiral multiplets in total is just MSSM ones, except the existence of those for three generations of right-handed neutrino and extra generations of MSSM Higgs pairs. Hierarchical Yukawa couplings can be obtained from the overlap integrals of wavefunctions localized in extra dimensions, allowing semi-realistic patterns of flavor structures for quarks and charged leptons. We also develop a systematic way to embed additional 7-branes into the configuration, those are sequestered from the visible sector toward a hidden sector model building.

hep-ph↗

arXiv2025

Domain Generalization of Pathological Image Segmentation by Patch-Level and WSI-Level Contrastive Learning

In this paper, we address domain shifts in pathological images by focusing on shifts within whole slide images~(WSIs), such as patient characteristics and tissue thickness, rather than shifts between hospitals. Traditional approaches rely on multi-hospital data, but data collection challenges often make this impractical. Therefore, the proposed domain generalization method captures and leverages intra-hospital domain shifts by clustering WSI-level features from non-tumor regions and treating these clusters as domains. To mitigate domain shift, we apply contrastive learning to reduce feature gaps between WSI pairs from different clusters. The proposed method introduces a two-stage contrastive learning approach WSI-level and patch-level contrastive learning to minimize these gaps effectively.

cs.CV↗

arXiv2025

Comprehensive Pathological Image Segmentation via Teacher Aggregation for Tumor Microenvironment Analysis

The tumor microenvironment (TME) plays a crucial role in cancer progression and treatment response, yet current methods for its comprehensive analysis in H&E-stained tissue slides face significant limitations in the diversity of tissue cell types and accuracy. Here, we present PAGET (Pathological image segmentation via AGgrEgated Teachers), a new knowledge distillation approach that integrates multiple segmentation models while considering the hierarchical nature of cell types in the TME. By leveraging a unique dataset created through immunohistochemical restaining techniques and existing segmentation models, PAGET enables simultaneous identification and classification of 14 key TME components. We demonstrate PAGET's ability to perform rapid, comprehensive TME segmentation across various tissue types and medical institutions, advancing the quantitative analysis of tumor microenvironments. This method represents a significant step forward in enhancing our understanding of cancer biology and supporting precise clinical decision-making from large-scale histopathology images.

cs.CV↗

arXiv2024

Roles of electric field/time-dependent Wilson line in toroidal compactification with or without magnetic fluxes

We discuss the effects of electric fields along compact directions within (supersymmetric) gauge theory in $\mathbb{R}^{1,3}\times{\mathbb T}^2 (\times{\mathbb T}^2\times{\mathbb T}^2)$. The electric field along compact directions is equivalent to time-dependent homogeneous configuration of Wilson line moduli, which would be relevant to physics in the early universe. In particular, we consider models with and without background magnetic fluxes, which lead to completely different effects of the electric field due to the difference in the Kaluza-Klein (KK) level structure. We show that, in the case without magnetic fluxes, the deceleration of KK momenta may cause non-perturbative KK particle production dubbed as the KK Schwinger effect, whereas in the case with magnetic flux such KK particle production does not take place but flavor structure of low energy effective theory may be affected.

hep-th↗

arXiv2024

One-loop vacuum energy in 10D super-Yang-Mills theory on magnetized tori with/without 4D N=1 supersymmetric completion

We compute the one-loop vacuum energy in 10 dimensional (10D) super Yang-Mills theory compactified on $\mathbb{R}^{1,3}\times (\mathbb{T}^2)^3$ in the presence of the Abelian magnetic fluxes. The regularization of infinite Kaluza-Klein (KK) sum is achieved by the use of Barnes $\zeta$-functions, which can be applied to the case with Landau level structure of KK mass spectrum. We apply the technique to two different models of the 10D super Yang-Mills theory: The one is to introduce the magnetic flux background to the 10D super Yang-Mills action. The other is to first embed 10D super Yang-Mills action into 4D $\mathcal{N}=1$ superspace and to introduce magnetic fluxes. The two model buildings result in tree level potential. The KK mass spectrum of two models look different and we compute the one-loop vacuum energy of each case. Nevertheless, both of the KK mass spectra have the Landau level structure and we can apply the regularization method in the same way. We show that despite the differences the resultant one-loop vacuum energy of both models coincide with each other.

hep-th↗

arXiv2023

Cluster Entropy: Active Domain Adaptation in Pathological Image Segmentation

The domain shift in pathological segmentation is an important problem, where a network trained by a source domain (collected at a specific hospital) does not work well in the target domain (from different hospitals) due to the different image features. Due to the problems of class imbalance and different class prior of pathology, typical unsupervised domain adaptation methods do not work well by aligning the distribution of source domain and target domain. In this paper, we propose a cluster entropy for selecting an effective whole slide image (WSI) that is used for semi-supervised domain adaptation. This approach can measure how the image features of the WSI cover the entire distribution of the target domain by calculating the entropy of each cluster and can significantly improve the performance of domain adaptation. Our approach achieved competitive results against the prior arts on datasets collected from two hospitals.

cs.CV↗

arXiv2023

Domain Adaptive Multiple Instance Learning for Instance-level Prediction of Pathological Images

Pathological image analysis is an important process for detecting abnormalities such as cancer from cell images. However, since the image size is generally very large, the cost of providing detailed annotations is high, which makes it difficult to apply machine learning techniques. One way to improve the performance of identifying abnormalities while keeping the annotation cost low is to use only labels for each slide, or to use information from another dataset that has already been labeled. However, such weak supervisory information often does not provide sufficient performance. In this paper, we proposed a new task setting to improve the classification performance of the target dataset without increasing annotation costs. And to solve this problem, we propose a pipeline that uses multiple instance learning (MIL) and domain adaptation (DA) methods. Furthermore, in order to combine the supervisory information of both methods effectively, we propose a method to create pseudo-labels with high confidence. We conducted experiments on the pathological image dataset we created for this study and showed that the proposed method significantly improves the classification performance compared to existing methods.

cs.CV↗

arXiv2023

Cluster-Guided Semi-Supervised Domain Adaptation for Imbalanced Medical Image Classification

Semi-supervised domain adaptation is a technique to build a classifier for a target domain by modifying a classifier in another (source) domain using many unlabeled samples and a small number of labeled samples from the target domain. In this paper, we develop a semi-supervised domain adaptation method, which has robustness to class-imbalanced situations, which are common in medical image classification tasks. For robustness, we propose a weakly-supervised clustering pipeline to obtain high-purity clusters and utilize the clusters in representation learning for domain adaptation. The proposed method showed state-of-the-art performance in the experiment using severely class-imbalanced pathological image patches.

cs.CV↗

arXiv2022

Achieving Transparency in Distributed Machine Learning with Explainable Data Collaboration

Transparency of Machine Learning models used for decision support in various industries becomes essential for ensuring their ethical use. To that end, feature attribution methods such as SHAP (SHapley Additive exPlanations) are widely used to explain the predictions of black-box machine learning models to customers and developers. However, a parallel trend has been to train machine learning models in collaboration with other data holders without accessing their data. Such models, trained over horizontally or vertically partitioned data, present a challenge for explainable AI because the explaining party may have a biased view of background data or a partial view of the feature space. As a result, explanations obtained from different participants of distributed machine learning might not be consistent with one another, undermining trust in the product. This paper presents an Explainable Data Collaboration Framework based on a model-agnostic additive feature attribution algorithm (KernelSHAP) and Data Collaboration method of privacy-preserving distributed machine learning. In particular, we present three algorithms for different scenarios of explainability in Data Collaboration and verify their consistency with experiments on open-access datasets. Our results demonstrated a significant (by at least a factor of 1.75) decrease in feature attribution discrepancies among the users of distributed machine learning.

cs.LG↗

arXiv2022

Non-readily identifiable data collaboration analysis for multiple datasets including personal information

Multi-source data fusion, in which multiple data sources are jointly analyzed to obtain improved information, has considerable research attention. For the datasets of multiple medical institutions, data confidentiality and cross-institutional communication are critical. In such cases, data collaboration (DC) analysis by sharing dimensionality-reduced intermediate representations without iterative cross-institutional communications may be appropriate. Identifiability of the shared data is essential when analyzing data including personal information. In this study, the identifiability of the DC analysis is investigated. The results reveals that the shared intermediate representations are readily identifiable to the original data for supervised learning. This study then proposes a non-readily identifiable DC analysis only sharing non-readily identifiable data for multiple medical datasets including personal information. The proposed method solves identifiability concerns based on a random sample permutation, the concept of interpretable DC analysis, and usage of functions that cannot be reconstructed. In numerical experiments on medical datasets, the proposed method exhibits a non-readily identifiability while maintaining a high recognition performance of the conventional DC analysis. For a hospital dataset, the proposed method exhibits a nine percentage point improvement regarding the recognition performance over the local analysis that uses only local dataset.

cs.LG↗

arXiv2020

Fate of domain walls in 5D gravitational theory with compact extra dimension

We pursue the time evolution of the domain walls in 5D gravitational theory with a compact extra dimension by numerical calculation. In order to avoid a kink-antikink pair that decays into the vacuum, we introduce a topological winding in the field space. In contrast to the case of non-gravitational theories, there is no static domain-wall solution in the setup. In the case that the minimal value of the potential is non-negative, we find that both the 3D space and the extra dimension will expand at late times if the initial value of the Hubble parameter is chosen as positive. The wall width almost remains constant during the evolution. In other cases, the extra dimension diverges and the 3D space shrinks to zero at a finite time.

hep-th↗

arXiv2020

Loop Fayet-Iliopoulos terms in $T^2/Z_2$ models: instability and moduli stabilization

We study Fayet-Iliopoulos (FI) terms of six-dimensional supersymmetric Abelian gauge theory compactified on a $T^2/Z_2$ orbifold. Such orbifold compactifications can lead to localized FI-terms and instability of bulk zero modes. We study 1-loop correction to FI-terms in more general geometry than the previous works. We find induced FI-terms depend on the complex structure of the compact space. We also find the complex structure of the torus can be stabilized at a specific value corresponding to a self-consistent supersymmetric minimum of the potential by such 1-loop corrections, which is applicable to the modulus stabilization.

hep-th↗

arXiv2019

Behaviors of two supersymmetry breaking scales in $\mathcal{N}=2$ supergravity

We study the supersymmetry breaking patterns in four-dimensional $\mathcal{N}=2$ gauged supergravity. The model contains multiple (Abelian) vector multiplets and a single hypermultiplet which parametrizes SO$(4,1)/{\rm{SO}}(4)$ coset. We derive the expressions of two gravitino masses under {\it{general}} gaugings and prepotential based on the embedding tensor formalism, and discuss their behaviors in some concrete models. Then we confirm that in a single vector multiplet case, the partial breaking always occurs when the third derivative of the prepotential exists at the vacuum, which is consistent with the result of Ref.~\cite{Antoniadis:2018blk}, but we can have several breaking patterns otherwise. The discussion is also generalized to the case of multiple vector multiplets, and we found that the full ($\mathcal{N}=0$) breaking occurs even if the third derivative of the prepotential is nontrivial.

hep-th↗

arXiv2019

Multi-Stage Pathological Image Classification using Semantic Segmentation

Histopathological image analysis is an essential process for the discovery of diseases such as cancer. However, it is challenging to train CNN on whole slide images (WSIs) of gigapixel resolution considering the available memory capacity. Most of the previous works divide high resolution WSIs into small image patches and separately input them into the model to classify it as a tumor or a normal tissue. However, patch-based classification uses only patch-scale local information but ignores the relationship between neighboring patches. If we consider the relationship of neighboring patches and global features, we can improve the classification performance. In this paper, we propose a new model structure combining the patch-based classification model and whole slide-scale segmentation model in order to improve the prediction performance of automatic pathological diagnosis. We extract patch features from the classification model and input them into the segmentation model to obtain a whole slide tumor probability heatmap. The classification model considers patch-scale local features, and the segmentation model can take global information into account. We also propose a new optimization method that retains gradient information and trains the model partially for end-to-end learning with limited GPU memory capacity. We apply our method to the tumor/normal prediction on WSIs and the classification performance is improved compared with the conventional patch-based method.

eess.IV↗

arXiv2019

Effects of localized μ-terms at the fixed points in magnetized orbifold models

We consider magnetized orbifolds, where the supersymmetric mass term for a pair of up- and down-type Higgs (super)fields, called $μ$-term, is localized at the orbifold fixed points, and study the effects on the zero-mode spectra. The zero-mode degeneracy to be identified as the generation in four-dimensional (4D) effective theories is determined by the magnetic fluxes. It is known that multiple Higgs zero-modes appear in general in magnetized orbifold models. We derive the analytic form of the $μ$-term matrix in the 4D effective theory generated by the localized sources on $T^2/Z_2$ orbifold fixed points, and find that this matrix can lead to a distinctive pattern of the eigenvalues that yields hierarchical $μ$-terms for the multiple Higgs fields. The lightest ones can be exponentially suppressed due to the localized wavefunctions of zero-modes determined by the fluxes, while the others are of the order of the compactification scale, which can provide a dynamical origin of the electroweak scale as well as a simultaneous decoupling of extra Higgs fields. We also show that a certain linear combination of the lightest Higgs fields could generate the observed mass ratios of down-type quarks through their Yukawa couplings determined by the wavefunctions.

hep-th↗

arXiv2019

Revisiting instabilities of $S^1/Z_2$ models with loop-induced Fayet-Iliopoulos terms

We study Fayet-Iliopoulos (FI) terms of 5-dimensional supersymmetric $U(1)$ gauge theory compactified on $S^1/Z_2$. In this model, loop diagrams including matter hypermultiplets and brane chiral multiplets induce FI-terms localized at the fixed points. Localized FI-terms lead instabilities of bulk modes. The form of the induced FI-terms strictly depends on wave function profiles of matter multiplets. It is a non-trivial question whether the vacuum of 1-loop corrected potential is stable under radiative corrections. We investigate this issue and it is found that the stable configuration is obtained when the bulk zero modes shield the brane charge completely.

hep-th↗

arXiv2019

Polonyi-Starobinsky supergravity with inflaton in a massive vector multiplet with DBI and FI terms

We propose the Starobinsky-type inflationary model in the matter-coupled $N=1$ four-dimensional supergravity with the massive vector multiplet that has inflaton (scalaron) and goldstino amongst its field components, whose action includes the Dirac-Born-Infeld-type kinetic term and the generalized (new) Fayet-Iliopoulos-type term, without gauging the R-symmetry. The $N=1$ chiral matter ("hidden sector") is described by the modified Polonyi model needed for spontaneous supersymmetry breaking after inflation. We compute the bosonic action and the scalar potential of the model, and show that it can accommodate the positive (observed) cosmological constant (as the dark energy) and the spontaneous supersymmetry breaking at high scale after the Starobinsky inflation.

hep-th↗

arXiv2019

Interpolation of partial and full supersymmetry breakings in $\cal{N} = 2$ supergravity

We discuss an $\cal{N}=2$ supergravity model that interpolates the full and the partial supersymmetry breakings. In particular, we find the conditions for an $\cal{N}=0$ Minkowski vacuum, which is continuously connected to the partial-breaking ($\cal{N}=1$ preserving) one. The model contains multiple (Abelian) vector multiplets and a single hypermultiplet, and is constructed by employing the embedding tensor technique. We compute the mass spectrum on the Minkowski vacuum, and find some non-trivial mass relations among the massive fields. Our model allows us to choose the two supersymmetry-breaking scales independently, and to discuss the cascade supersymmetry breaking for the applications to particle phenomenology and cosmology.

hep-th↗