日本語のページは こちら

Akifumi Okuno1,2,3,†
1Inst. of Stat. Math.  2SOKENDAI U.  3RIKEN AIP
okuno@ism.ac.jp

Biography
He studies statistical machine learning and mathematical statistics as an assistant professor of the Institute of Statistical Mathematics (located in Tachikawa, Tokyo, Japan). He got his Ph.D. degree in Informatics from Kyoto University (Sep. 2020) under the supervision of Prof. Hidetoshi Shimodaira. Also see his curriculum vitae, Researchmap, and YouTube channel providing video lecture materials. Overview of his study is also summarized in https://okuno.net/overview.pdf.
Keywords: Statistical Machine Learning, Neural Network, Nonparametric Theories, Scientific Applications, Representation Learning, Graph Embdding, Correlation Analysis, Robust Estimation.
His research outline (youtube)

#Papers

#Preprint

  1. Akifumi Okuno* and Shotaro Yagishita*

    Outlier-Robust Neural Network Training: Efficient Optimization of Transformed Trimmed Loss with Variation Regularization

    arXiv: October 2024

    submitted

    https://doi.org/10.48550/arXiv.2308.02293

    oknakfm/ARTL

    論文解説 (和文)

    Two authors (Okuno and Yagishita) contributed equally to this work. This manuscript is a complete rewrite of the earlier preprint HOVR (arXiv:2308.02293v2), which was released in August 2023 (unpublished and not intended for future publication), with the addition of new authors and the introduction of new techniques. Please note that arXiv versions v3 and beyond refer to this manuscript as an update to the earlier versions (v1 and v2).

  2. Akifumi Okuno

    An integrated perspective of robustness in regression through the lens of the bias-variance trade-off

    arXiv: July 2024

    in preparation for submission

    https://doi.org/10.48550/arXiv.2407.10418

    in Japanese

    oknakfm/RBVT

  3. Akifumi Okuno

    A multivariate adaptation of direct kernel estimation of density ratio

    arXiv: November 2023

    in preparation for resubmission

    https://doi.org/10.48550/arXiv.2311.12380

    in Japanese

    oknakfm/MKDRE

  4. Akifumi Okuno and Kohei Hattori

    A Greedy and Optimistic Approach to Clustering with a Specified Uncertainty of Covariates

    arXiv: April 2022

    submitted

    https://doi.org/10.48550/arXiv.2204.08205

    Video in English

    oknakfm/GOC

#Journal

  1. Akifumi Okuno

    Minimizing robust density power-based divergences for general parametric density models

    Annals of the Institute of Statistical Mathematics

    2024

    https://doi.org/10.1007/s10463-024-00906-9

    Video in English and in Japanese

    oknakfm/DPD

    R package: oknakfm/sgdpd, パッケージ解説動画 (日本語)

    論文解説 (和文)

  2. Akifumi Okuno and Kazuharu Harada

    An interpretable neural network-based non-proportional odds model for ordinal regression

    Journal of Computational and Graphical Statistics

    2024

    https://doi.org/10.1080/10618600.2024.2321208

    Video in Japanese

    oknakfm/N3POM

    論文解説 (和文)

  3. Akifumi Okuno, Yuya Morishita, and Yoh-ichi Mototake

    Autoregressive with Slack Time Series Model for Forecasting a Partially-Observed Dynamical Time Series

    IEEE Access

    2024

    https://doi.org/10.1109/ACCESS.2024.3365724

    Video in Japanese

    oknakfm/ARS

  4. Akifumi Okuno and Masaaki Imaizumi

    Minimax Analysis for Inverse Risk in Nonparametric Planer Invertible Regression

    Electronic Journal of Statistics

    2024

    https://doi.org/10.1214/23-EJS2202

    Video in English

    oknakfm/NPIR

    論文解説 (和文)

  5. Akifumi Okuno and Keisuke Yano

    A generalization gap estimation for overparameterized models via the Langevin functional variance

    Journal of Computational and Graphical Statistics

    2023

    https://doi.org/10.1080/10618600.2023.2197488

    Video in English and in Japanese

    論文解説 (和文)

  6. Kohei Hattori, Akifumi Okuno, and Ian U. Roederer

    Finding r-II sibling stars in the Milky Way with the Greedy Optimistic Clustering algorithm

    Astrophysical Journal

    2023

    https://doi.org/10.3847/1538-4357/acb93b

  7. Akifumi Okuno and Keisuke Yano

    Dependence of variance on covariate design in nonparametric link regression

    Statistics and Probability Letters

    2023

    https://doi.org/10.1016/j.spl.2022.109716

    Video in Japanese

  8. Akifumi Okuno and Hidetoshi Shimodaira

    Hyperlink Regression via Bregman Divergence

    Neural Networks

    2020

    https://doi.org/10.1016/j.neunet.2020.03.026

#Conference

  1. Akifumi Okuno and Hidetoshi Shimodaira

    Extrapolation Towards Imaginary 0-Nearest Neighbour and Its Improved Convergence Rate

    Advances in Neural Information Processing Systems (NeurIPS)

    2020

    https://papers.nips.cc/paper/2020/hash/f9028faec74be6ec9b852b0a542e2f39-Abstract.html

    in Japanese

    論文解説 (和文)

  2. Geewook Kim, Akifumi Okuno, Kazuki Fukui, and Hidetoshi Shimodaira

    Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities

    Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI)

    2019

    https://www.ijcai.org/proceedings/2019/699

  3. Akifumi Okuno and Hidetoshi Shimodaira

    Robust Graph Embedding with Noisy Link Weights

    22nd International Conference on Artificial Intelligence and Statistics (AISTATS)

    2019

    http://proceedings.mlr.press/v89/okuno19b.html

  4. Akifumi Okuno, Geewook Kim, and Hidetoshi Shimodaira

    Graph Embedding with Shifted Inner Product Similarity and Its Improved Approximation Capability

    22nd International Conference on Artificial Intelligence and Statistics (AISTATS)

    2019

    http://proceedings.mlr.press/v89/okuno19a.html

  5. Akifumi Okuno, Tetsuya Hada, and Hidetoshi Shimodaira

    A probabilistic framework for multi-view feature learning with many-to-many associations via neural networks

    35th International Conference on Machine Learning (ICML)

    2018

    http://proceedings.mlr.press/v80/okuno18a.html

  6. Kazuki Fukui, Akifumi Okuno, and Hidetoshi Shimodaira

    Image and tag retrieval by leveraging image-group links with multi-domain graph embeddings

    Proceedings of the 2016 IEEE International Conference on Image Processing (ICIP)

    2016

    https://doi.org/10.1109/ICIP.2016.7532351

#MinorWorks

  1. Akifumi Okuno, Takumi Kodahara, and Makoto Sasaki

    Hierarchical Clustering of Modes in Numerical Turbulence Fields

    Plasma and Fusion Research: Rapid Communications

    2024

    Accepted

    oknakfm/HCFS

  2. Akifumi Okuno and Hidetoshi Shimodaira

    On representation power of neural network-based graph embedding and beyond

    ICML 2018 workshop on Theoretical Foundations and Applications of Deep Generative Models (TADGM)

    2018

    https://sites.google.com/view/tadgm/accepted-papers

  3. Tetsuya Hada, Akifumi Okuno, and Hidetoshi Shimodaira

    Deep Multi-view Representation Learning Based on Adaptive Weighted Similarity

    International Workshop on Symbolic-Neural Learning (SNL)

    2017

    https://www.ttic.edu/SNL2017/schedule.htm

#Shelved

*No plans for future publication. Left as is.
  1. Akifumi Okuno

    A stochastic optimization approach to train non-linear neural networks with regularization of higher-order total variation

    arXiv: August 2023

    It has been incorporated into the newer ARTL manuscript and will not be published in the future. Left as is

    https://arxiv.org/abs/2308.02293v2

    Video in English and in Japanese

    oknakfm/HOVR

    This manuscript has been completely rewritten as a new manuscript, ARTL, with the addition of new authors and the introduction of new techniques. Namely, the HOVR manuscript (arXiv v1 and v2) has been incorporated into the ARTL manuscript (arXiv versions v3 and beyond).

  2. Ruixing Cao*, Akifumi Okuno*, Kei Nakagawa, and Hidetoshi Shimodaira

    Improving Nonparametric Classification via Local Radial Regression with an Application to Stock Prediction

    arXiv: December 2021

    Left as is

    https://doi.org/10.48550/arXiv.2112.13951

    in Japanese

    First two authors (Cao and Okuno) contributed equally to this work.

  3. Morihiro Mizutani, Akifumi Okuno, Geewook Kim, and Hidetoshi Shimodaira

    Stochastic Neighbor Embedding of Multimodal Relational Data for Image-Text Simultaneous Visualization

    arXiv: May 2020

    Left as is

    https://doi.org/10.48550/arXiv.2005.00670

#論文解説

  1. 奥野彰文, 下平英寿

    論文解説:仮想的な0近傍法による高次バイアス補正

    Jxiv

    2024

    https://doi.org/10.51094/jxiv.945

  2. 奥野彰文, 柳下翔太郎

    論文解説:外れ値にロバストなニューラルネットの学習

    Jxiv

    2024

    https://doi.org/10.51094/jxiv.928

  3. 奥野彰文

    論文解説:一般の確率モデルでの冪密度ダイバージェンス最小化

    Jxiv

    2024

    https://doi.org/10.51094/jxiv.642

  4. 奥野彰文,今泉允聡

    論文解説:可逆関数推定の難しさ - 生成モデルを念頭に

    Jxiv

    2024

    https://doi.org/10.51094/jxiv.616

  5. 奥野彰文,原田和治

    論文解説:順序回帰における柔軟性とドメイン制約のトレードオフ

    Jxiv

    2023

    https://doi.org/10.51094/jxiv.549

  6. 奥野彰文,矢野恵佑

    論文解説:WAIC による過剰パラメータモデルの汎化誤差推定

    Jxiv

    2023

    https://doi.org/10.51094/jxiv.537

#国内会議論文/レター

  1. Ruixing Cao, 田中卓磨, 奥野彰文, 下平英寿

    マルチスケールk-近傍法における回帰関数および損失関数の検討

    人工知能学会全国大会論文集

    2021

    https://doi.org/10.11517/pjsai.JSAI2021.0_1G4GS2c01

  2. 田中卓磨, 奥野彰文, 下平英寿

    マルチスケールk-近傍法による画像のExtreme Multi-Label分類

    人工知能学会全国大会論文集

    2021

    https://doi.org/10.11517/pjsai.JSAI2021.0_3G1GS2g03

  3. 水谷守裕, 奥野彰文, 福井一輝, Kim Geewook, 下平英寿

    グラフと近傍グラフの確率的同時埋め込みを用いたマルチモーダル関連性データの可視化

    人工知能学会全国大会論文集

    2020

    https://doi.org/10.11517/pjsai.JSAI2020.0_2P5GS301

  4. Kim Geewook, 奥野彰文, 下平英寿

    擬ユークリッド空間への単語埋め込み

    言語処理学会第25回年次大会論文集

    2019

    http://www.anlp.jp/proceedings/annual_meeting/2019/pdf_dir/P7-4.pdf

  5. 福井一輝, 奥野彰文, 下平英寿

    マッチング相関分析を用いた画像-マルチタグ間の相互検索

    電子情報通信学会和文論文誌D: 研究速報(レター)

    2016

    https://doi.org/10.14923/transinfj.2015IUL0005

#Presentations

  1. Akifumi Okuno

    A stochastic optimization approach to minimize robust density power-based divergences for general parametric density models

    ISI-ISM-ISSAS meeting

    Kolkata, India

    Dec. 2023

  2. Akifumi Okuno, Ruixing Cao, Kei Nakagawa, and Hidetoshi Shimodaira

    Optimal nonparametric classification via radial distance

    CMStatistics

    Berlin, Germany

    Dec. 2023

  3. Akifumi Okuno

    Estimation with integral-based loss functions

    International Symposium on Recent Advances in Theories and Methodologies for Large Complex Data

    Tsukuba, Japan

    Dec. 2023

  4. Akifumi Okuno and Masaaki Imaizumi

    Minimax Analysis for Inverse Risk in Nonparametric Invertible Regression

    The 6th RIKEN-IMI-ISM-NUS-ZIB-MODAL-NHR Workshop on Advances in Classical and Quantum Algorithms for Optimization and Machine Learning

    Fukuoka, Japan

    Sep. 2022

  5. Akifumi Okuno and Kohei Hattori

    A Greedy and Optimistic Approach to Clustering with a Specified Uncertainty of Covariates

    JJSM2022 JSS-KSS-CSA Joint Session (3):Machine Learning

    Online

    Sep. 2022

  6. Akifumi Okuno and Keisuke Yano

    A generalization gap estimation for overparameterized models via the Langevin functional variance

    Workshop on Functional Inference and Machine Intelligence

    Online

    Mar. 2022

  7. Akifumi Okuno and Masaaki Imaizumi

    Estimation of Invertible Functions

    ISI-ISM-ISSAS meeting

    Online

    Jan. 2022

  8. Masaaki Imaizumi and Akifumi Okuno

    Minimax Analysis for Inverse Risk in Nonparametric Planer Invertible Regression

    CMStatistics

    Online

    Dec. 2021

  9. Keisuke Yano and Akifumi Okuno

    On estimating generalization gaps via the functional variance in overparameterized model

    CMStatistics

    Online

    Dec. 2021

  10. Akifumi Okuno and Keisuke Yano

    Nonparametric Link Regression and Its Theoretical Properties

    EcoSta

    Online

    Jun. 2021

  11. Akifumi Okuno and Hidetoshi Shimodaira

    Extrapolation Towards Imaginary 0-Nearest Neighbour and Its Improved Convergence Rate

    NeurIPS meetup in Japan

    Online

    Dec. 2020

  12. Akifumi Okuno, Hidetoshi Shimodaira, and Geewook Kim

    Bregman Hyperlink Regression and Its Expressive Power

    ACML 2019 Workshop on Statistics and Machine Learning Researchers in Japan

    Nagoya, Japan

    Nov. 2019

  13. Akifumi Okuno and Hidetoshi Shimodaira

    Hyperlink Regression via Bregman Divergence

    RIKEN-AIP workshop

    Genoa, Italy

    Sep. 2019

  14. Akifumi Okuno, Hidetoshi Shimodaira, and Geewook Kim

    Graph Embedding with Shifted Inner Product Similarity and Its Improved Approximation Capability

    Workshop on Functional Inference and Machine Intelligence

    Tokyo, Japan

    Mar. 2019

  15. Akifumi Okuno and Hidetoshi Shimodaira

    Leveraging local data structure for multi-view analysis with many-to-many associations

    Conference of the International Federation of Classification Societies (IFCS)

    Tokyo, Japan

    Aug. 2017

  16. Akifumi Okuno and Hidetoshi Shimodaira

    Statistical consistency of multi-view correlation analysis with many-to-many association

    Joint Statistical Meeting (JSM)

    Baltimore, USA

    Aug. 2017

  17. Akifumi Okuno and Hidetoshi Shimodaira

    Robust Multi-view Graph Embedding

    International Conference on Robust Statistics (ICoRS)

    Wollongong, Australia

    Jul. 2017

  18. Akifumi Okuno and Hidetoshi Shimodaira

    Robust cross-domain matching: Analyzing multi-domain data vectors under mismatched association

    Machine Learning Summer School

    Kyoto, Japan

    Aug. 2015

*Presentations for the papers listed in #Papers#Conference are excluded. Please also see Researchmap for the remaining presentations in Japanese.

#Awards

  1. Ogawa Research Award, Japan Statistical Society, 2024.
  2. Excellent Presentation Award (in the top 3~6/200), IBIS2023, 2023.
  3. Excellent Presentation Award (in the top 5~7/238), IBIS2019, 2019.
  4. Excellent Research Award, ICT-13@Kyoto University, 2019.

#Social

@public_aokn | https://okuno.net/ja