Research activity information

• Risk-Averse Best Arm Set Identification with Fixed Budget and Fixed Confidence.
  Shunta Nonaga; Koji Tabata; Yuta Mizuno; Tamiki Komatsuzaki
  CoRR, abs/2506.22253, Jun. 2025
  Scientific journal
• Comparative Analysis of Reinforcement Learning Algorithms for Finding Reaction Pathways: Insights from a Large Benchmark Data Set
  Yoshihiro Matsumura; Koji Tabata; Tamiki Komatsuzaki
  Journal of Chemical Theory and Computation, American Chemical Society (ACS), 19 Mar. 2025
  Scientific journal
• Multi-armed bandit algorithm for sequential experiments of molecular properties with dynamic feature selection.
  Md Menhazul Abedin; Koji Tabata; Yoshihiro Matsumura; Tamiki Komatsuzaki
  The Journal of chemical physics, 161, 1, 07 Jul. 2024, [International Magazine]
  English, Scientific journal, Sequential optimization is one of the promising approaches in identifying the optimal candidate(s) (molecules, reactants, drugs, etc.) with desired properties (reaction yield, selectivity, efficacy, etc.) from a large set of potential candidates, while minimizing the number of experiments required. However, the high dimensionality of the feature space (e.g., molecular descriptors) makes it often difficult to utilize the relevant features during the process of updating the set of candidates to be examined. In this article, we developed a new sequential optimization algorithm for molecular problems based on reinforcement learning, multi-armed linear bandit framework, and online, dynamic feature selections in which relevant molecular descriptors are updated along with the experiments. We also designed a stopping condition aimed to guarantee the reliability of the chosen candidate from the dataset pool. The developed algorithm was examined by comparing with Bayesian optimization (BO), using two synthetic datasets and two real datasets in which one dataset includes hydration free energy of molecules and another one includes a free energy difference between enantiomer products in chemical reaction. We found that the dynamic feature selection in representing the desired properties along the experiments provides a better performance (e.g., time required to find the best candidate and stop the experiment) as the overall trend and that our multi-armed linear bandit approach with a dynamic feature selection scheme outperforms the standard BO with fixed feature variables. The comparison of our algorithm to BO with dynamic feature selection is also addressed.
• Flow zoometry of Drosophila
  Walker Peterson; Joshua Arenson; Soichiro Hata; Laura Kacenauskaite; Tsubasa Kobayashi; Takuya Otsuka; Hanqing Wang; Yayoi Wada; Kotaro Hiramatsu; Zhikai He; Jean-Emmanuel Clement; Chenqi Zhang; Chenglang Hu; Phillip McCann; Hayato Kanazawa; Yuzuki Nagasaka; Hiroyuki Uechi; Yuh Watanabe; Ryodai Yamamura; Mika Hayashi; Yuta Nakagawa; Kangrui Huang; Hiroshi Kanno; Yuqi Zhou; Tianben Ding; Maik Herbig; Shimpei Makino; Shunta Nonaga; Ryosuke Takami; Oguz Kanca; Koji Tabata; Satoshi Amaya; Kotaro Furusawa; Kenichi Ishii; Kazuo Emoto; Fumihito Arai; Ross Cagan; Dino Di Carlo; Tatsushi Igaki; Erina Kuranaga; Shinya Yamamoto; Hugo J Bellen; Tamiki Komatsuzaki; Masahiro Sonoshita; Keisuke Goda
  Cold Spring Harbor Laboratory, 05 Apr. 2024
  ABSTRACT
  
  Drosophilaserves as a highly valuable model organism across numerous fields including genetics, immunology, neuroscience, cancer biology, and developmental biology. Central toDrosophila-based biological research is the ability to perform comprehensive genetic or chemical screens. However, this research is often limited by its dependence on laborious manual handling and analysis, making it prone to human error and difficult to discern statistically significant or rare events amid the noise of individual variations resulting from genetic and environmental factors. In this article we present flow zoometry, a whole-animal equivalent of flow cytometry for large-scale, individual-level, high-content screening ofDrosophila. Our flow zoometer automatically clears the tissues ofDrosophila melanogaster, captures three-dimensional (3D) multi-color fluorescence tomograms of single flies with single-cell volumetric resolution at an unprecedented throughput of over 1,000 animals within 48 hours (24 hr for clearing; 24 hr for imaging), and performs AI-enhanced data-driven analysis – a task that would traditionally take months or years with manual techniques. To demonstrate its broad applications, we employed the flow zoometer in various laborious screening assays, including those in toxicology, genotyping, and tumor screening. Flow zoometry represents a pivotal evolution in high-throughput screening technology: previously from molecules to cells, now from cells to whole animals. This advancement serves as a foundational platform for “statistical spatial biology”, to improve empirical precision and enable serendipitous discoveries across various fields of biology.
• On-the-fly Raman microscopy guaranteeing the accuracy of discrimination.
  Koji Tabata; Hiroyuki Kawagoe; J Nicholas Taylor; Kentaro Mochizuki; Toshiki Kubo; Jean-Emmanuel Clement; Yasuaki Kumamoto; Yoshinori Harada; Atsuyoshi Nakamura; Katsumasa Fujita; Tamiki Komatsuzaki
  Proceedings of the National Academy of Sciences of the United States of America, 121, 12, e2304866121, 19 Mar. 2024, [International Magazine]
  English, Scientific journal, Accelerating the measurement for discrimination of samples, such as classification of cell phenotype, is crucial when faced with significant time and cost constraints. Spontaneous Raman microscopy offers label-free, rich chemical information but suffers from long acquisition time due to extremely small scattering cross-sections. One possible approach to accelerate the measurement is by measuring necessary parts with a suitable number of illumination points. However, how to design these points during measurement remains a challenge. To address this, we developed an imaging technique based on a reinforcement learning in machine learning (ML). This ML approach adaptively feeds back "optimal" illumination pattern during the measurement to detect the existence of specific characteristics of interest, allowing faster measurements while guaranteeing discrimination accuracy. Using a set of Raman images of human follicular thyroid and follicular thyroid carcinoma cells, we showed that our technique requires 3,333 to 31,683 times smaller number of illuminations for discriminating the phenotypes than raster scanning. To quantitatively evaluate the number of illuminations depending on the requisite discrimination accuracy, we prepared a set of polymer bead mixture samples to model anomalous and normal tissues. We then applied a home-built programmable-illumination microscope equipped with our algorithm, and confirmed that the system can discriminate the sample conditions with 104 to 4,350 times smaller number of illuminations compared to standard point illumination Raman microscopy. The proposed algorithm can be applied to other types of microscopy that can control measurement condition on the fly, offering an approach for the acceleration of accurate measurements in various applications including medical diagnosis.
• Gaussian process classification bandits.
  Tatsuya Hayashi; Naoki Ito; Koji Tabata; Atsuyoshi Nakamura; Katsumasa Fujita; Yoshinori Harada; Tamiki Komatsuzaki
  Pattern Recognit., 149, 110224, 110224, 2024
  Scientific journal
• Differentiability of cell types enhanced by detrending a non-homogeneous pattern in a line-illumination Raman microscope.
  Abdul Halim Bhuiyan; Jean-Emmanuel Clément; Zannatul Ferdous; Kentaro Mochizuki; Koji Tabata; James Nicholas Taylor; Yasuaki Kumamoto; Yoshinori Harada; Thomas Bocklitz; Katsumasa Fujita; Tamiki Komatsuzaki
  The Analyst, 148, 15, 3574, 3583, 05 Jul. 2023, [International Magazine]
  English, Scientific journal, A line illumination Raman microscope extracts the underlying spatial and spectral information of a sample, typically a few hundred times faster than raster scanning. This makes it possible to measure a wide range of biological samples such as cells and tissues - that only allow modest intensity illumination to prevent potential damage - within feasible time frame. However, a non-uniform intensity distribution of laser line illumination may induce some artifacts in the data and lower the accuracy of machine learning models trained to predict sample class membership. Here, using cancerous and normal human thyroid follicular epithelial cell lines, FTC-133 and Nthy-ori 3-1 lines, whose Raman spectral difference is not so large, we show that the standard pre-processing of spectral analyses widely used for raster scanning microscopes introduced some artifacts. To address this issue, we proposed a detrending scheme based on random forest regression, a nonparametric model-free machine learning algorithm, combined with a position-dependent wavenumber calibration scheme along the illumination line. It was shown that the detrending scheme minimizes the artifactual biases arising from non-uniform laser sources and significantly enhances the differentiability of the sample states, i.e., cancerous or normal epithelial cells, compared to the standard pre-processing scheme.
• 単調増加制約のあるレベルセット推定
  田畑 公次; 中村 篤祥; 高見 亮佑; Joshua Arenson; 和田 弥生; Walker Peterson; 合田 圭介; 園下 将大; 小松崎 民樹
  人工知能学会研究会資料 人工知能基本問題研究会, 124, 25, 30, 一般社団法人 人工知能学会, 06 Mar. 2023
  Japanese
• Raman imaging of rat non‐alcoholic fatty liver tissues reveals distinct biomolecular states
  Khalifa Mohammad Helal; Harsono Cahyadi; J. Nicholas Taylor; Akira Okajima; Koji Tabata; Yasuaki Kumamoto; Kentaro Mochizuki; Yoshito Itoh; Tetsuro Takamatsu; Hideo Tanaka; Katsumasa Fujita; Tamiki Komatsuzaki; Yoshinori Harada
  FEBS Letters, Wiley, 17 Feb. 2023, [Peer-reviewed]
  Scientific journal
• Posterior Tracking Algorithm for Classification Bandits.
  Koji Tabata; Junpei Komiyama; Atsuyoshi Nakamura; Tamiki Komatsuzaki
  AISTATS, 10994, 11022, PMLR, 2023
  International conference proceedings
• Lipid droplet accumulation and adipophilin expression in follicular thyroid carcinoma.
  Michiyo Hayakawa; J Nicholas Taylor; Ryuta Nakao; Kentaro Mochizuki; Yuki Sawai; Kosuke Hashimoto; Koji Tabata; Yasuaki Kumamoto; Katsumasa Fujita; Eiichi Konishi; Shigeru Hirano; Hideo Tanaka; Tamiki Komatsuzaki; Yoshinori Harada
  Biochemical and biophysical research communications, 640, 192, 201, 05 Dec. 2022, [International Magazine]
  English, Scientific journal, Follicular neoplasms of the thyroid include follicular thyroid carcinoma (FTC) and follicular thyroid adenoma (FTA). However, the differences in cytological findings between FTC and FTA remain undetermined. Here, we aimed to evaluate the accumulation of lipid droplets (LDs) and the expression of adipophilin (perilipin 2/ADRP/ADFP), a known LD marker, in cultured FTC cells. We also immunohistochemically compared adipophilin expression in the FTC and FTA of resected human thyroid tissues. Cultured FTC (FTC-133 and RO82W-1) possessed increased populations of LDs compared to thyroid follicular epithelial (Nthy-ori 3-1) cells. In vitro treatment with phosphatidylinositol-3-kinase (PI3K)/Akt/mammalian target of rapamycin (mTOR) signaling inhibitors (LY294002, MK2206, and rapamycin) in FTC-133 cells downregulated the PI3K/Akt/mTOR/sterol regulatory element-binding protein 1 (SREBP1) signaling pathway, resulting in a significant reduction in LD accumulation. SREBP1 is a master transcription factor that controls lipid metabolism. Fluorescence immunocytochemistry revealed adipophilin expression in the LDs of FTC-133 cells. Immunohistochemical analysis of surgically resected human thyroid tissues revealed significantly increased expression of adipophilin in FTC compared with FTA and adjacent non-tumorous thyroid epithelia. Taken together, LDs and adipophilin were abundant in cultured FTC; the evaluation of adipophilin expression can help distinguish FTC from FTA in surgical specimens.
• Gaussian Process Classification Bandits.
  Tatsuya Hayashi; Naoki Ito; Koji Tabata; Atsuyoshi Nakamura; Katsumasa Fujita; Yoshinori Harada; Tamiki Komatsuzaki
  CoRR, abs/2212.13157, 2022
  Scientific journal
• Classification Bandits: Classification Using Expected Rewards as Imperfect Discriminators.
  Koji Tabata; Atsuyoshi Nakamura; Tamiki Komatsuzaki
  Trends and Applications in Knowledge Discovery and Data Mining - PAKDD 2021 Workshops, WSPA, MLMEIN, SDPRA, DARAI, and AI4EPT, 57, 69, Springer, 2021
  International conference proceedings
• A bad arm existence checking problem: How to utilize asymmetric problem structure?
  Koji Tabata; Atsuyoshi Nakamura; Junya Honda; Tamiki Komatsuzaki
  Mach. Learn., 109, 2, 327, 372, 2020
  Scientific journal
• Raman spectroscopic histology using machine learning for nonalcoholic fatty liver disease.
  Khalifa Mohammad Helal; James Nicholas Taylor; Harsono Cahyadi; Akira Okajima; Koji Tabata; Yoshito Itoh; Hideo Tanaka; Katsumasa Fujita; Yoshinori Harada; Tamiki Komatsuzaki
  FEBS letters, 593, 18, 2535, 2544, Sep. 2019, [International Magazine]
  English, Scientific journal, Histopathology requires the expertise of specialists to diagnose morphological features of cells and tissues. Raman imaging can provide additional biochemical information to benefit histological disease diagnosis. Using a dietary model of nonalcoholic fatty liver disease in rats, we combine Raman imaging with machine learning and information theory to evaluate cellular-level information in liver tissue samples. After increasing signal-to-noise ratio in the Raman images through superpixel segmentation, we extract biochemically distinct regions within liver tissues, allowing for quantification of characteristic biochemical components such as vitamin A and lipids. Armed with microscopic information about the biochemical composition of the liver tissues, we group tissues having similar composition, providing a descriptor enabling inference of tissue states, contributing valuable information to histological inspection.
• A Bad Arm Existence Checking Problem.
  Koji Tabata; Atsuyoshi Nakamura; Junya Honda; Tamiki Komatsuzaki
  CoRR, abs/1901.11200, 2019
  Scientific journal
• Feature selection as Monte-Carlo Search in Growing Single Rooted Directed Acyclic Graph by Best Leaf Identification.
  Aurélien Pélissier; Atsuyoshi Nakamura; Koji Tabata
  Proceedings of the 2019 SIAM International Conference on Data Mining(SDM), 450, 458, SIAM, 2019
  International conference proceedings
• 機械学習による非アルコール性脂肪性肝疾患モデルのラマン分光分析
  原田 義規; Cahyadi Harsono; Helal Khalifa; 岡嶋 亮; 田畑 公次; Taylor Nick; 熊本 康昭; 高松 哲郎; 小松崎 民樹; 田中 秀央
  日本病理学会会誌, 107, 1, 368, 368, (一社)日本病理学会, Apr. 2018
  Japanese
• Bad Arm Existence Checking Algorithms with Small Sample Complexity
  TABATA Koji; NAKAMURA Atsuyoshi; HONDA Junya; KOMATSUZAKI Tamiki
  JSAI Technical Report, SIG-FPAI, 106, 16, The Japanese Society for Artificial Intelligence, 07 Mar. 2018
  Japanese
• Feature selection as Monte-Carlo Search in Growing Single Rooted Directed Acyclic Graph by Best Leaf Identification.
  Aurélien Pélissier; Atsuyoshi Nakamura; Koji Tabata
  CoRR, abs/1811.07531, 2018
  Scientific journal
• An Efficient Approximate Algorithm for the 1-Median Problem on a Graph
  TABATA Koji; NAKAMURA Atsuyoshi; KUDO Mineichi
  IEICE Transactions on Information and Systems, 100, 5, 994, 1002, The Institute of Electronics, Information and Communication Engineers, 2017
  English,

  We propose a heuristic approximation algorithm for the 1-median problem. The 1-median problem is the problem of finding a vertex with the highest closeness centrality. Starting from a randomly selected vertex, our algorithm repeats to find a vertex with higher closeness centrality by approximately calculating closeness centrality of each vertex using simpler spanning subgraphs, which are called k-neighbor dense shortest path graphs with shortcuts. According to our experimental results using real networks with more than 10,000 vertices, our algorithm is more than 100 times faster than the exhaustive search and more than 20 times faster than the state-of-the-art approximation algorithm using annotated information to the vertices while the solutions output by our algorithm have higher approximation ratio.

• An Algorithm for Influence Maximization in a Two-Terminal Series Parallel Graph and its Application to a Real Network
  Koji Tabata; Atsuyoshi Nakamura; Mineichi Kudo
  DISCOVERY SCIENCE, DS 2015, 9356, 275, 283, 2015, [Peer-reviewed]
  English, International conference proceedings

• Rapid diagnosis using sequential Raman measurements
  田畑公次; 川越寛之; TAYLOR J.; 望月健太郎; 久保俊貴; CLEMENT Jean-Emmanuel; 熊本康昭; 原田義規; 中村篤祥; 藤田克昌; 小松崎民樹, 人工知能学会全国大会論文集(Web), 38th, 2024
• 機械学習による非アルコール性脂肪性肝疾患モデルのラマン分光分析
  原田義規; CAHYADI Harsono; HELAL Khalifa; 岡嶋亮; 田畑公次; TAYLOR Nick; 熊本康昭; 高松哲郎; 小松崎民樹; 田中秀央, 日本病理学会会誌, 107, 1, 368, 368, 26 Apr. 2018
  (一社)日本病理学会, Japanese
• Fast Approximation Algorithm for the 1-Median Problem
  田畑 公次; 中村 篤祥; 工藤 峰一, 人工知能基本問題研究会, 86, 77, 82, 09 Aug. 2012
  人工知能学会, Japanese
• Fast Algorithm for Finding a Graph Node with High Closeness Centrality
  TABATA Koji; NAKAMURA Atsuyoshi; KUDO Mineichi, IEICE technical report. Theoretical foundations of Computing, 111, 256, 7, 14, 14 Oct. 2011
  The Closeness Centrality is one of centrality measures of a node in a graph. It is calculated as the reciprocal of the sum of distances to all other nodes. In this paper, we propose a fast approximate algorithm that finds the node to maximize its closeness centrality in an undirected graph. Given a node v, the algorithm repeatedly find a node with higher closeness centrality by making use of a shortest path tree of the previous node. According to out experiment, our algorithm can find a node with almost maximum closeness centrality with high probability., The Institute of Electronics, Information and Communication Engineers, Japanese

• 大学院共通授業科目（一般科目）：自然科学・応用科学, 2024年, 修士課程, 大学院共通科目
• 大学院共通授業科目（一般科目）：自然科学・応用科学, 2024年, 修士課程, 大学院共通科目
• 大学院共通授業科目（一般科目）：自然科学・応用科学, 2024年, 修士課程, 大学院共通科目
• 大学院共通授業科目（一般科目）：自然科学・応用科学, 2024年, 修士課程, 大学院共通科目
• 大学院共通授業科目（一般科目）：自然科学・応用科学, 2024年, 修士課程, 大学院共通科目
• 数理科学概説, 2024年, 修士課程, 理学院
• 数学総合講義Ⅰ, 2024年, 学士課程, 理学部
• 微分積分学Ⅰ, 2024年, 学士課程, 全学教育
• 微分積分学Ⅱ, 2024年, 学士課程, 全学教育

• 漸近的最適かつ実用的な純粋探索バンディット方策の開発
  科学研究費助成事業
  01 Apr. 2024 - 31 Mar. 2029
  中村 篤祥; 田畑 公次; 畑埜 晃平; 寺本 央
  日本学術振興会, 基盤研究(A), 北海道大学, 24H00685
• Construction of Multi-Armed Bandit Methods for Practical Applications of Pure Exploration Problems
  Grants-in-Aid for Scientific Research
  01 Apr. 2024 - 31 Mar. 2027
  田畑 公次
  Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (C), Hokkaido University, 24K15064
• Acceleration of the Development of Organic Reactions Based on the Fusion of Automatic Synthesis Robots and Information Science
  Grants-in-Aid for Scientific Research
  01 Apr. 2021 - 31 Mar. 2024
  長田 裕也; 水野 雄太; 田畑 公次; 辻 信弥; 小林 正人
  有機合成研究において反応条件の最適化は極めて重要であり、研究遂行のためには多大な労力と時間を費やし続けている。本研究は、自動合成ロボットと量子化学計算によって得られる分子記述子を用いた強化学習を用いることで、反応条件の最適化を飛躍的に加速させることを目的としている。
  ２０２１年度の研究においては、自動合成ロボットにおける予備的有機合成実験を実施し、さらに実験結果の自動測定に取り組んだ。初期検討として、熱的フィスゲン環化付加反応と縮合剤を用いたアミド化反応に関する検討に取り組んだ。遷移金属触媒を用いずにフィスゲン環化付加反応を行った場合、1,4-付加体と1,5-付加体の混合物が得られ、これらの生成比は基質の構造に依存する。あらかじめ量子化学計算によって生成比の予測を行い、続いて自動合成ロボットを用いた合成実験と生成比の決定を行うことで両者の比較を行い、反応モデルの改良を行うことで、良い精度で生成比の予測を行う方法の開発に成功した。
  また、縮合剤を用いたアミド化反応に関する検討では、種々のカルボン酸類とアミン類からアミド化合物を合成し、アミド類の構造と超臨界流体クロマトグラフィーでの保持時間の関係について検討を行った。超臨界流体クロマトグラフィーではポリブチレンテレフタレートがコーティングされたシリカゲルを固定相として用いたカラムを使用することで、迅速な分離分析を行うことができることを見出した。現在、アミド類の構造とその保持時間の相関について分子記述子を用いた解析を進めており、未知のアミド化合物の分析条件の推定が可能になるものと期待している。
  Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (B), Hokkaido University, 21H01924
• バンディット問題の方策の実用化のための理論の深化
  科学研究費助成事業
  01 Apr. 2019 - 31 Mar. 2023
  中村 篤祥; 田畑 公次; 工藤 峰一
  昨年度から始めた分類バンディット問題のアルゴリズムの開発において、k-腕設定のトンプ ソンサンプリングをベースとしたアルゴリズム、および実験的な性能評価を国際学会のワークショップで発表した。更に指定した信頼度で正しく判定するためのサンプル数の理論的評価を始めた。連続腕設定に拡張した問題においては、ガウス過程を事前分布とする方式において、全腕の正負(報酬の期待値が閾値以上か未満か）を答える既存法より少ないサンプル数の上界が得られるアルゴリズムを開発した。また、正腕の割合の推定値から分類結果を予測し、その予測が正しい場合に分類の精度が上がる方の腕を引く方式を考案し、有効性を実験により確かめた。ラマン分光による細胞診断の実データにも適用し、グリット分割によるk腕設定のトンプソンサンプリングより、ガウス過程を仮定した連続腕設定の提案方式の方が少ないサンプル数（計測数）で正しく判定できることを確認した。
  小ノイズ敵対的バンディットの研究においては、以前の成果である「ノイズフリーバンディット 問題」を小さなノイズを許す問題に拡張し、それに有効なアルゴリズムを開発することを目指している。ノイズフ リー条件(1回も誤らない腕が存在するという条件)を「誤る回数が高々k回の腕が存在する」という条件に緩めた問題定式化で研究を進めている。昨年度に引き続き、その問題設定 の下、腕の数が3本以上でk>2の条件でも動作するもっと一般的なアルゴリズムの開発を進めた。
  精度効率保証大規模探索の研究においては、昨年度に引き続き属性選択アルゴリズム[Aurelien, Nakamura, Tabata 2019]のアルゴリズムの探索木の拡張法の検討を行った。
  日本学術振興会, 基盤研究(B), 北海道大学, 19H04161
• Development of efficient multi-armed bandit algorithm for good arm identification and its application
  Grants-in-Aid for Scientific Research
  01 Apr. 2018 - 31 Mar. 2022
  Koji Tabata
  The purpose of this research is development and application of efficient methods to identify good arms, whose expected reward is larger then a given threshold, under the multi-armed bandit setting which is a model of trade-off between knowledge exploration and exploitation. Here, an efficient method means that it can identify the good arm as few samples as possible.
  We confirmed that our proposed methods have better performance then exiting method. We have also developed a prototype diagnostic device using the algorithm developed in this research.
  Japan Society for the Promotion of Science, Grant-in-Aid for Early-Career Scientists, Hokkaido University, 18K18099

• 計測制御装置、分光計測装置、及び計測制御方法
  Patent right, 中村 篤祥; 田畑 公次; 小松崎 民樹; 藤田 克昌, 国立大学法人北海道大学, 国立大学法人大阪大学
  特願2018-244664, 27 Dec. 2018
  特開2020-106370, 09 Jul. 2020
  202003014497985620