An optimal transport and information geometric framework for infinite-dimensional Gaussian measures and Gaussian processes (II)

2026年7月28日(火) 15:00 - 16:15

Minh Ha Quang (理化学研究所 革新知能統合研究センター (AIP) 不完全情報学習チーム 上級研究員)

Divergences between probability distributions play a crucial role in many areas of probability theory, statistics, machine learning, and their applications. While a large part of the literature is focused on divergences between finite-dimensional distributions, there is a growing body of work on infinite-dimensional distances/divergences, which are motivated by applications in functional data analysis, Bayesian inverse problems, and functional Bayesian neural networks, among others. In this lecture, we present an overview of recent results on some of the most important divergences being studied, including the Kullback-Leibler, Renyi, and Geometric Jensen-Shannon divergences. We discuss the many challenges that arise in the infinite-dimensional setting, e.g. the lack of a natural reference measure such as the Lebesgue measure and the fact that many functions such as determinants and logarithm are only well-defined in specific settings. In particular, in the setting of Gaussian measures on infinite-dimensional Hilbert spaces, the closed form expressions for the above divergences are only generalizable to equivalent Gaussian measures. We present the resolution to the above challenges via the geometrical framework of positive definite unitized (or regularized) trace class and Hilbert-Schmidt operators, including the Alpha and Alpha-Beta Log-Determinant divergences. Using this framework and the methodology of reproducing kernel Hilbert spaces (RKHS), we furthermore obtain consistent finite-dimensional  approximations of the above divergences in the Gaussian process setting, with dimensional-independent sample complexities. The resulting numerical algorithms can be readily employed in practical applications. We shall also discuss the generalization of the above classical divergences above to the quantum setting, namely the Quantum Jensen-Shannon divergence between quantum states, defined in terms of the von Neumann and Tsallis entropies, from finite to infinite-dimensional settings.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

An optimal transport and information geometric framework for infinite-dimensional Gaussian measures and Gaussian processes (I)

2026年7月28日(火) 13:30 - 14:45

Minh Ha Quang (理化学研究所 革新知能統合研究センター (AIP) 不完全情報学習チーム 上級研究員)

Optimal transport (OT) and information geometry (IG) have been attracting much research attention in various fields, in particular machine learning and statistics. In this lecture, we present results on the generalization of IG and OT distances for finite-dimensional Gaussian measures to the setting of infinite-dimensional Gaussian measures and Gaussian processes. Our focus is on the Entropic Regularization of the 2-Wasserstein distance and the generalization of the Fisher-Rao Riemannian metric and related quantities. In both settings, regularization leads to many desirable theoretical properties, including in particular dimension-independent convergence and sample complexity. The mathematical formulation involves the interplay of IG and OT with Gaussian processes and the methodology of reproducing kernel Hilbert spaces (RKHS). All of the presented formulations admit closed form expressions that can be efficiently computed and applied practically. The mathematical formulations will be illustrated with numerical experiments on Gaussian processes.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

Center-vortex condensation and monopole condensation in 4d gapped phases

2026年7月27日(月) 14:00 - 15:30

林 優依 (京都大学 基礎物理学研究所 学振特別研究員PD)

Two well-known scenarios for quark confinement are center-vortex proliferation and monopole condensation. We consider gauge-invariant criteria for center-vortex condensation and monopole condensation in terms of Z(N) 1-form symmetry. The condensation of a soliton can be characterized by the non-suppression of the partition function with a proper twisted boundary condition, and we utilize this idea for these criteria. With these definitions, we show that gapped phases with the center-vortex condensation necessarily exhibit the monopole condensation.

会場: セミナー室 (359号室) 3階 359号室とZoomのハイブリッド開催

イベント公式言語: 英語

Some instances where topological illustration induced new mathematics

2026年7月24日(金) 16:30 - 18:00

Sofia Lambropoulou (Professor, School of Applied Mathematical and Physical Sciences, National Technical University of Athens, Greece)

We shall present instances from generalized knot theory, braid theory and their interactions, where illustration promoted understanding and inspired new mathematics. The first instance addresses a question of V.F.R. Jones whether one can make analogous constructions to the (2-variable) Jones polynomial using other braid groups and other types of Hecke algebras. The second instance addresses the question of formulating braid equivalences, analogous to the Markov theorem for classical braids, in settings where we may not even have available algebraic structures for the related braids. The third instance is about the theory of bonded knots and bonded knotoids used for modelling proteins.

会場: via Zoom / セミナー室 (359号室)

イベント公式言語: 英語

Unraveling the very early universe with black holes, boson stars, and cannibal stars

2026年7月24日(金) 14:00 - 16:00

小林 洸 (Associate Professor, International School for Advanced Studies (SISSA), Italy)

According to the standard picture of cosmology, the rich structure of our universe began to form roughly 50,000 years after the big bang. In this talk I will explore the possibility that cosmic structures could also have formed in the extremely early universe, within a fraction of a second after inflation. I will show how this early structure formation can give rise to compact objects, including exotic stars and primordial black holes. These relics provide powerful probes of the first instants of cosmic history, especially the reheating epoch, and may even act as seeds for cosmological phase transitions. Note: This seminar is jointly organized by the iTHEMS-phys Study Group and the iTHEMS-ABBL Joint Astro Study Group.

会場: セミナー室 (359号室) / via Zoom

イベント公式言語: 英語

The decision intelligence of humans and machines

2026年7月24日(金) 10:30 - 11:30

Petter Holme (Professor, Department of Computer Science, Aalto University, Finland)

The event has been rescheduled from July 22 to July 24. To understand our near-future of artificial intelligence firmly integrated into many levels of social life, a challenge is to understand the differences and similarities between human and AI decision-making. In controlled laboratory settings assessing risk and uncertainty, LLMs demonstrate superhuman efficiency but fundamentally diverge from human behavior through a rigid hyper-rationality and an inability to disengage from obsolete strategies. However, when applied to messy, real-world dilemmas "in the wild," these models pivot to function as highly effective "satisficers". Human subjects consistently prefer this artificial counsel over human peer advice, noting its ability to carefully balance emotional context with logical constraints while actively reducing anxiety and regret. Ultimately, this synthesis shows that while AI can offer near-optimal laboratory performance and therapeutic impact in daily life, they also have a distinct lack of behavioral plasticity that we need to account for in models of the future.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

When Is Collective Intelligence a Lottery? Toward a Physics of Bounded Agent Societies

2026年7月22日(水) 16:00 - 17:30

田中 秀宣 (Group Leader, CBS-NTT Physics of Intelligence Program, Center for Brain Science, Harvard University, USA)

Multi-agent LLM systems can reach agreement quickly, but agreement alone does not reveal whether a group has integrated evidence, amplified a bias, or merely locked in the luck of early samples. In this talk, I will develop a physics-style approach to this problem through two synthetic games. In reward-free naming games, populations form conventions through mutual in-context learning: agents treat one another's sampled outputs as evidence, so early fluctuations compound into consensus. A minimal model, Quantized Simplex Gossip (QSG), identifies this regime as memetic drift and predicts scaling laws and a crossover from lottery-like drift to bias-driven selection. I will then introduce the Flag Game, in which a hidden country flag provides verifiable ground truth but each agent sees only a private crop of it. Here, bounded agents must balance private evidence against social input. We find rich phenomenologies, where adding agents can help or hurt, large groups can polarize, and social-awareness prompting, model diversity, and organizational structure all reshape collective performance. Extending QSG with grounded evidence and model-specific update rules explains these effects. More broadly, these studies frame an LLM society as a network of networks, neural networks coupled through social interaction, and outline a route toward a multiscale physics of interacting AI agents, linking model-internal representations, agent-level decision mechanisms, and population-level social dynamics.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

Workshop on Discrete & Continuous Aspects of Reaction-Diffusion in Pattern Formation (DiCoRD2026)

2026年7月22日(水) - 24日(金)

富安 亮子 (九州大学 マス・フォア・インダストリ研究所 教授)
佐藤 純 (金沢大学 教授)
末松 信彦 (明治大学 総合数理学部 専任教授)
西浦 廉政 (北海道大学 名誉教授)
リッカルド・ムオロ (理化学研究所 数理創造研究センター (iTHEMS) 数理基礎部門 基礎科学特別研究員)
Jonathan Dawes (Professor, University of Bath, UK)
Henrik Weyer (Postdoctoral Scholar, University of California, Santa Barbara, USA)
加藤 譲 (公立はこだて未来大学 システム情報科学部 複雑系知能学科 准教授)
小澤 歩 (海洋研究開発機構 (JAMSTEC) Young Research Fellow)
杉本 貴則 (関西大学 システム理工学部 准教授)
義永 那津人 (公立はこだて未来大学 システム情報科学部 教授)
Jens Rademacher (Professor, University of Hamburg, Germany)
深尾 武史 (龍谷大学 先端理工学部 教授)
田中 吉太郎 (公立はこだて未来大学 システム情報科学部 准教授)
石原 秀至 (東京大学 特任准教授)
石井 宙志 (北海道大学 電子科学研究所 助教授)
渡辺 毅 (長野大学 准教授)
アントワーヌ・ディエズ (理化学研究所 数理創造研究センター (iTHEMS) 数理展開部門 数学応用研究チーム 研究員)

会場: via Zoom / 明治大学 中野キャンパス 高層棟6F 研究セミナー室3

イベント公式言語: 英語

Entanglement suppression for ΩΩ scattering

2026年7月17日(金) 15:00 - 16:30

曽根 克佳 (東京都立大学 大学院理学研究科 博士課程)

The S-matrix describing the scattering process can be expressed in terms of projection operators onto the allowed spin–flavor channels and the corresponding phase shifts. Using the entanglement entropy in the spin space of the two-particle state, one can define the entanglement power, which quantifies the ability of the S-matrix to generate entanglement in the system. By investigating the conditions under which the entanglement power of the S-matrix is minimized, namely, the conditions for entanglement suppression, one can derive relations among the phase shifts in different spin–flavor channels. Furthermore, by comparing these relations with the interaction Lagrangian, one can identify the underlying symmetries [1,2]. In this work, we apply the entanglement suppression framework to two-baryon scattering involving spin-3/2 baryons in the flavor decuplet [3]. Lattice QCD calculations have shown that the spin-0 ΩΩ system exhibits scattering close to the unitary limit. Combining this result with the relation between the phase shifts obtained from entanglement suppression, we discuss the scattering behavior of the spin-2 ΩΩchannel.

会場: 研究本館 4階 445-447号室 (メイン会場) / via Zoom

イベント公式言語: 英語

第33回 MACSコロキウム

2026年7月17日(金) 14:45 - 18:00

成瀬 元 (京都大学 大学院理学研究科 地球惑星科学専攻 教授)
村瀬 洋介 (理化学研究所 数理創造研究センター (iTHEMS) 数理展開部門 数理社会科学チーム チームディレクター)

14:45-15:00 ティータイムディスカッション 15:00-16:00 成瀬 元（京都大学大学院理学研究科 地球惑星科学専攻 教授）「地層は何を覚えているのか？―機械学習で探る過去の地球環境―」 16:15-17:15 村瀬 洋介（理化学研究所 数理創造研究センター チームディレクター）「社会における協力行動の数理：直接互恵と間接互恵から見る協力の進化」 17:15-18:00 継続討論会

会場: 京都大学 理学研究科セミナーハウス（建物配置図（北部構内）【10】の建物）

イベント公式言語: 日本語

A first step towards Non-Archimedean Geometric Quantization

2026年7月17日(金) 14:00 - 15:30

後藤 慶太 (理化学研究所 数理創造研究センター (iTHEMS) 数理基礎部門 基礎科学特別研究員)

Calabi--Yau manifolds have long attracted interest from both mathematics and physics, particularly in the context of mirror symmetry, and form an important class of compact Kähler manifolds. A compact Kähler manifold is Calabi--Yau if and only if it admits a Ricci-flat Kähler metric, which we shall call a CY metric. Such a metric is highly analytic in nature, as it is given as the solution to a second-order PDE on the manifold, namely the complex Monge--Ampère equation. When the Calabi--Yau manifold is a complex projective variety, one algebraic approach to understanding this analytically defined CY metric is to approximate it by algebraically defined metrics called balanced metrics. This framework was initiated by Donaldson and is now known as geometric quantization. In this talk, following the spirit of this theory, we consider a non-Archimedean analogue of this approximation theory. More precisely, for a non-Archimedean analytic space associated with a maximally degenerating family of Calabi--Yau manifolds, we study the approximation of the NACY metric, a non-Archimedean analogue of the CY metric, by algebraically defined metrics. In particular, we introduce NA balanced metrics, which are expected to provide such an approximation, and explain that, for totally degenerating families of abelian varieties, NA balanced metrics indeed approximate the NACY metric.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

Mathematical Application Research Team Meeting #17

2026年7月17日(金) 10:30 - 11:45

Hugo Koubbi (Ph.D. Student, CEREMADE, Université Paris Dauphine, France)

Mathematical Application Research Team is delighted to welcome Hugo Koubbi from CEREMADE, Université Paris Dauphine, for an upcoming team meeting. We warmly invite everyone to join us to hear his talk: Title: Mean-field Transformers Abstract: Since their introduction, Transformers have achieved unprecedented success across a wide range of deep-learning tasks. In this talk, we present a recent line of work that interprets Transformer architectures as interacting particle systems and studies their continuum, or mean-field, limits. By considering idealized attention dynamics on the sphere, we establish connections between Transformers, Wasserstein gradient flows, and synchronization models such as the Kuramoto model. We describe a global clustering phenomenon in which tokens eventually synchronize, following long-lived metastable regimes characterized by the coexistence of several clusters. Finally, we explain how recent work extends these ideas beyond idealized models and reveals analogous phenomena in more realistic Transformer architectures at initialization.

会場: via Zoom / セミナー室 (359号室) 3階 359号室

イベント公式言語: 英語

Artificial Intelligence in New Physics Electroweak Phase Transition Studies

2026年7月16日(木) 15:00 - 16:00

Yang Zhang (Professor, Henan Normal University, China)

The study of electroweak phase transitions in BSM involves complex numerical calculations, large parameter spaces, and the integration of multiple computational tools. In this talk, I will review recent developments in applying artificial intelligence to new physics phase transition studies. First, I will discuss how machine learning methods can accelerate electroweak phase transition studies, including efficient evaluations of phase transition dynamics, such as bounce action calculations, and the exploration of detectable parameter regions for gravitational-wave searches. Then, I will introduce the emerging role of AI agents in scientific workflows, including automated model construction, effective potential generation, and parameter scans. These developments illustrate how AI can transform traditional computational pipelines and provide new possibilities for future high-energy physics research.

会場: via Zoom

イベント公式言語: 英語

Visual system and (social) behavior in zebrafish

2026年7月16日(木) 13:00 - 14:00

久保 郁 (理化学研究所 脳神経科学研究センター (CBS) 知覚運動統合機構研究チーム チームディレクター)

Animals make decisions about their actions using sensory information from the external world. Our lab investigates how the brain processes visual information and generates appropriate behavior using the vertebrate model organism, zebrafish. Our research has uncovered the neural circuits required for processing optic flow, a visual cue that animals use to estimate their own motion. We employ a diverse range of techniques, including behavioral tracking, live imaging of neural activity, and molecular and genetic characterization of neuronal cell types. More recently, our research has focused on social behavior, in which groups of animals collectively generate behavioral decisions. In this seminar, I will present our recent work investigating behavioral contagion in zebrafish.

会場: セミナー室 (359号室) 3階 359号室とZoomのハイブリッド開催

イベント公式言語: 英語

Joint Seminar on Cosmology

2026年7月13日(月) 13:30 - 18:00

神田 行宏 行宏 (東京大学 宇宙線研究所 (ICRR) 特任研究員)
岡松 郁弥 (日本大学 文理学部 物理学科 助手)
佐野 文哉 (東京大学 大学院総合文化研究科 JSPS特別研究員PD)

The Joint Seminar is a collaborative seminar series organized by universities and research institutes in and around Tokyo. It is held once every one or two months, with the venue rotating among the participating institutions. At each meeting, we have around three talks and open, informal discussions. After the seminar, participants are also welcome to join an informal social gathering.

会場: 大河内記念ホール

イベント公式言語: 英語

How did we come to be? — Particle Physics for the Next Decades —

2026年7月10日(金) 15:30 - 17:00

村山 斉 (東京大学 カブリ数物連携宇宙研究機構 (Kavli IPMU) 教授 / Professor, Department of Physics, University of California, Berkeley, USA)

Particle Physics is a study of the smallest and the biggest to uncover the fundamental laws that govern the universe. In recent years, both the United States and Europe have been through long-range planning processes. The future plans worldwide include the studies of (1) neutrinos that may have saved us from a complete annihilation, (2) the Higgs boson that keeps us in one piece, (3) dark matter that assembled us from the primordial soup, (4) inflation that created the macroscopic universe, and (5) the exploration of unknown particles and forces. It requires development of mind-boggling technologies.

会場: 大河内記念ホール (メイン会場) / via Zoom

イベント公式言語: 英語

What determines accuracy in matrix projection models?

2026年7月10日(金) 14:30 - 15:30

Richard Shefferson (東京大学 大学院総合文化研究科 教授)

Matrix projection models (MPMs) have grown in complexity as ecologists have sought to include more factors that may influence population size and structure. However, some studies suggest that MPMs may lead to predictions inaccurate enough as to question their overall utility. I used long-term (21-36 year) demographic datasets on 6 herbaceous perennial species to examine and compare the ability of MPMs with different structural characteristics to predict future population size and structure. In absolute terms, almost all models performed poorly. In relative terms, density-dependent, ahistorical stage-based models with simple life histories and fewer stages were most successful in predicting population size. My results indicate that MPMs and IPMs are typically poor predictors of absolute population size and structure, but, when constructed properly, can still be used as useful qualitative predictors of population change.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

Multi-field Inflation with spectator Axions

2026年7月9日(木) 14:00 - 16:00

Diederik Roest (Professor, Van Swinderen Institute for Particle Physics and Gravity, University of Groningen, Netherlands)

We will review the inflationary paradigm, with a focus on multi-field inflation. We then propose a generic mechanism in which a light axion spectator reshapes inflationary observables through purely gravitational multi‑field dynamics. In this scenario, the axion is frozen during inflation and starts rolling towards the end, inducing a turn in field space and transient tachyonic phases of the isocurvature mode. This generates a nearly scale‑invariant enhancement of the curvature power spectrum, suppressing the tensor‑to‑scalar ratio and shifting the scalar tilt to a weighted combination of adiabatic and entropic tilts at horizon crossing. We show that these effects can reconcile otherwise disfavored inflaton potentials with current CMB constraints, and predict order-one non-Gaussianities.

会場: セミナー室 (359号室) (メイン会場) / via Zoom

イベント公式言語: 英語

Sampling Lattice Gauge Theories with Gauge-Equivariant Neural Networks and Diffusion Models

2026年7月8日(水) 14:00 - 15:30

Andreas Ipp (Senior Scientist, Institute for Theoretical Physics, Technische Universität Wien, Austria)
Thomas Ranner (Project Assistant (FWF), Institute for Theoretical Physics, Technische Universität Wien, Austria)

Simulating non-Abelian lattice gauge theories using classical Markov chain Monte Carlo (MCMC) methods is often bottlenecked by critical slowing down and topological freezing. Incorporating physical symmetries into machine learning architectures offers a powerful way towards overcoming these computational limitations. In this talk, we present an overview of how gauge-equivariant neural networks can enhance lattice simulations. In the first part, we introduce Lattice Gauge Equivariant Convolutional Neural Networks (L-CNNs), which build local gauge symmetry directly into the network structure. We discuss their ability to learn Wilson loops and highlight their versatility across different applications, such as approximating effective actions. In the second part, we focus on their application in generative modeling: gauge-equivariant diffusion models. Using a Metropolis-adjusted annealed Langevin scheme, these models are designed to generate uncorrelated field configurations. We demonstrate accurate sampling of 2D U(2) and SU(2) theories, showcase recent advancements in scaling up to 4D SU(3) pure gauge theory, and show that these models extrapolate remarkably well to larger lattices and inverse couplings beyond their training regime.

会場: 研究本館 (メイン会場) / via Zoom

イベント公式言語: 英語

iTHEMS-FQSP joint seminar “Aspects of Tripartite Haar Random States”

2026年7月8日(水) 10:30 - 12:00

吉田 紅 (Research Faculty Senior Faculty, Perimeter Institute for Theoretical Physics, Canada / 理化学研究所 基礎量子科学研究プログラム 客員主管研究員)

Randomness plays central roles in understanding strongly entangled quantum systems. The foundational result is Page’s theorem: a bipartite Haar random state is nearly maximally entangled. In this talk, I will ask what happens when the system is divided into three parts. We show that tripartite Haar random states have a very different structure: when each subsystem contains fewer than half of the total qubits, no EPR-like bipartite entanglement can be distilled between any pair by local unitaries or local operations. I will discuss several consequences of this observation, including its implications for quantum error correction, complementary recovery, connected entanglement wedges in AdS/CFT, and possible baby-universe degrees of freedom.

会場: 大河内記念ホール (メイン会場) / via Zoom

イベント公式言語: 英語