Institution profile

Institute of Statistical Mathematics

Academic institutionasia · jp
Official website
Research library147linked papers
Opportunities0open roles
Selected work

Representative Papers

Augmented Shuffle Protocols for Accurate and Robust Frequency Estimation under Differential Privacy

Apr 10, 2025

Existing differentially private (DP) shuffling models face two critical challenges: poor robustness against local data poisoning attacks—especially under small ε—and vulnerability to privacy budget inflation when the data collector colludes with users. This paper proposes an enhanced shuffling framework that achieves pure ε-DP frequency estimation while provably resisting collusion. Our method introduces a universal protocol requiring no local noise injection, integrates randomized sampling and virtual data injection, and employs an asymmetric two-sided geometric distribution for virtual counts—ensuring strict ε-DP and effectively mitigating poisoning effects. We provide formal theoretical proofs establishing both ε-DP compliance and robustness against adversarial poisoning. Empirical evaluation demonstrates that, under identical privacy budgets, our approach improves estimation accuracy by 15–30% over state-of-the-art methods, achieving a superior balance among utility, computational efficiency, and rigorous privacy guarantees.

1 citationsRead paper

Distance-Based Tree-Sliced Wasserstein Distance

Mar 14, 2025

Existing Tree-Sliced Wasserstein on Systems of Lines (TSW-SL) methods rely solely on support point locations while ignoring the projection domain, and their fixed partitioning mappings lack Euclidean invariance, hindering preservation of high-dimensional topological structure. Method: We propose Distance-aware tree-sliced Wasserstein (Db-TSW), which introduces a class of distance-sensitive generalized splitting mappings that explicitly encode full positional information of the input metric, ensuring Euclidean invariance. We establish a Radon-transform-based theoretical framework, proving Db-TSW’s injectivity and metric validity, and design GPU-efficient tree sampling and distance computation mechanisms. Contribution/Results: Experiments demonstrate that Db-TSW significantly outperforms mainstream Sliced Wasserstein variants across diverse tasks—achieving notable accuracy gains—while maintaining linear time complexity and low computational overhead.

1 citationsRead paper

Minimum Copula Divergence for Robust Estimation

Feb 24, 2025

To address the lack of robustness of classical maximum likelihood estimation (MLE) under model misspecification and heavy-tailed data, this paper proposes a copula-only robust estimation framework. The method introduces a family of α/β/γ-divergences specifically tailored to copulas and constructs the minimum copula divergence estimator (MCDE), which directly minimizes the divergence between a parametric copula and the empirical copula—bypassing marginal distribution modeling entirely. Theoretically, we establish bounded influence functions for MCDE under Archimedean and elliptical copula families, ensuring strong robustness. Numerical experiments demonstrate that MCDE consistently outperforms MLE under model misspecification, contamination by extreme values, and diverse dependence structures, while maintaining adaptability and computational feasibility.

1 citationsRead paper

Testing the Homogeneity of Two Proportions for Correlated Bilateral Data via the Clayton Copula

Feb 01, 2025

In clinical trials—particularly ophthalmology—homogeneity testing for bilateral proportion data is commonly constrained by pre-specified, inflexible dependence structures (e.g., independence, perfect positive/negative dependence), limiting interpretability and adaptability. This paper introduces the Clayton copula—a flexible, interpretable tool for modeling asymmetric lower-tail dependence—into bilateral proportion homogeneity testing for the first time, thereby eliminating reliance on a priori dependence assumptions. We propose three Clayton copula–based test statistics and rigorously evaluate them via Monte Carlo simulation, demonstrating well-controlled Type I error rates and superior statistical power. Furthermore, we validate the robustness and practical utility of our approach on two real-world ophthalmologic datasets. This work establishes a theoretically rigorous, computationally feasible, and clinically meaningful testing paradigm for bilateral proportion data, advancing both methodological foundations and applied biostatistical practice.

1 citationsRead paper

Solving stochastic climate-economy models: A deep least-squares Monte Carlo approach

Aug 19, 2024arXiv.org

Traditional grid-based dynamic programming becomes computationally intractable for high-dimensional stochastic climate-economy models—such as the five-dimensional DICE model—due to the curse of dimensionality arising from numerous state variables and stochastic shocks. Method: This paper proposes a deep learning–enhanced Least-Squares Monte Carlo (LSMC) method, wherein deep neural networks replace the conventional linear regression in LSMC to approximate value functions and policy mappings. Contribution/Results: To our knowledge, this is the first approach enabling end-to-end dynamic optimal control for coupled multi-source uncertainties—including climate sensitivity and abatement cost heterogeneity—in high-dimensional stochastic settings. The method substantially improves both computational efficiency and solution accuracy for stochastic optimal control problems. Applied to the full five-dimensional stochastic DICE model, it successfully derives robust carbon tax trajectories and green investment strategies. This work establishes a scalable, high-precision computational paradigm for quantitative climate policy evaluation.

1 citationsRead paper
Recent publications

Latest Papers