Our Research
Recent Topics
- Information-Geometric Visualization and Evaluation of Large Language Models via Log-Likelihood Vectors and KL Divergence
- Geometric Analysis of Word Embeddings with Independent Component Analysis
- Information-Geometric Interpretation of Embedding Norms and KL Divergence
- Semantic Shift Detection via Optimal Transport and Temporal Changes in Embeddings
Ongoing Topics
- Model Selection and Information Criteria
- Multiscale Bootstrap Resampling and Selective Inference
- Statistical Inference under Covariate Shift (Term coined in Shimodaira JSPI 2000)
- Statistical Hypothesis Testing in Phylogenetic Inference
- Representation Learning for Graphs and Knowledge Bases
- Statistical Inference for Growing Complex Networks
Software
- Our github site is https://github.com/shimo-lab
- Universal-Geometry-with-ICA [paper] Discovering Universal Geometry in Embeddings with ICA (arxiv) (EMNLP2023 main)
- hoc [paper] Understanding Higher-Order Correlations Among Semantic Components in Embeddings (arxiv) (EMNLP2024 main)
- Axis-Tour [paper] Axis Tour: Word Tour Determines the Order of Axes in ICA-transformed Embeddings (arxiv) (EMNLP2024 findings)
- Statistical Analysis of Complex Networks: PAFit [paper] PAFit: an R Package for Estimating Preferential Attachment and Node Fitness in Temporal Complex Networks (arXiv) (Journal of Statistical Software 2020)
- Confidence level of hierarchical clustering: pvclust [paper] Pvclust: an R package for assessing the uncertainty in hierarchical clustering
- Confidence level of phylogenetic analysis: CONSEL [paper] CONSEL: for assessing the confidence of phylogenetic tree selection
- Multiscale bootstrap: scaleboot
- Natural Language Processing (older): senbei, kadingir, SCNE
- Graph embedding (older): SIPS
- Old Software Site
Paper Citations
- Hidetoshi Shimodaira (Google Scholar)
- Hidetoshi Shimodaira (Web of Science)
- Hidetoshi Shimodaira (Scopus)
- Hidetoshi Shimodaira (Semantic Scholar)
Recent papers
- Mapping 1,000+ Language Models via the Log-Likelihood Vector (arxiv)
- Quantifying Lexical Semantic Shift via Unbalanced Optimal Transport (arxiv)
- Norm of Mean Contextualized Embeddings Determines their Variance (arxiv) (COLING 2025)
- Revisiting Cosine Similarity via Normalized ICA-transformed Embeddings (arxiv) (COLING 2025)
- Zipfian Whitening (arxiv) (NeurIPS2024)
- Understanding Higher-Order Correlations Among Semantic Components in Embeddings (arxiv) (EMNLP2024 main)
- Axis Tour: Word Tour Determines the Order of Axes in ICA-transformed Embeddings (arxiv) (EMNLP2024 findings)
- Block-Diagonal Orthogonal Relation and Matrix Entity for Knowledge Graph Embedding (arxiv) (EMNLP2024 findings)
- Shimo Lab at “Discharge Me!”: Discharge Summarization by Prompt-Driven Concatenation of Electronic Health Record Sections (arxiv) (BioNLP-WS 2024) (ACL 2024 BioNLP Workshop) (BioNLP ACL’24 Shared Task on Streamlining Discharge Documentation)
- Predicting Drug-Gene Relations via Analogy Tasks with Word Embeddings (arxiv)
- Knowledge Sanitization of Large Language Models (arxiv)
- Discovering Universal Geometry in Embeddings with ICA (arxiv) (EMNLP2023 main)
- Norm of word embedding encodes information gain (arxiv) (EMNLP2023 main)
- Improving word mover’s distance by leveraging self-attention matrix (arxiv) (EMNLP2023 findings)
- 3D Rotation and Translation for Hyperbolic Knowledge Graph Embedding (arxiv) (EACL2024)
- Selective inference after feature selection via multiscale bootstrap (AISM 2022) (arXiv)
- A Hypergraph Approach for Estimating Growth Mechanisms of Complex Networks (IEEE Access 2022)
- Improving Nonparametric Classification via Local Radial Regression with an Application to Stock Prediction (ArXiv 2021)
- Revisiting Additive Compositionality: AND, OR and NOT Operations with Word Embeddings (arXiv 2021) (ACL-IJCNLP 2021 Student Research Workshop)
- Nonparametric estimation of the preferential attachment function from one network snapshot (arXiv 2021) (Journal of Complex Networks 2021)
- Extrapolation Towards Imaginary 0-Nearest Neighbour and Its Improved Convergence Rate (arXiv 2020) (NeurIPS 2020 accepted)
- Stochastic Neighbor Embedding of Multimodal Relational Data for Image-Text Simultaneous Visualization (arXiv 2020)
- Hyperlink Regression via Bregman Divergence (arXiv 2019) (Neural Networks)
- Joint Estimation of the Non-parametric Transitivity and Preferential Attachment Functions in Scientific Co-authorship Networks (Journal of Informetrics) (arXiv 2019)
- PAFit: an R Package for Estimating Preferential Attachment and Node Fitness in Temporal Complex Networks (arXiv) (Journal of Statistical Software 2020)
- More Powerful Selective Kernel Tests for Feature Selection (arViv 2019) (AISTATS 2020) [Joint work at Yamada team]
- Selective inference after feature selection via multiscale bootstrap (arXiv 2019 updated) (old title is Selective inference after variable selection via multiscale bootstrap)
- Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities (arXiv 2019) (IJCAI 2019) (Talk Slide)
- Segmentation-free compositional n-gram embedding (arXiv) (NAACL-HLT 2019 accepted papers) (software)
- Robust Graph Embedding with Noisy Link Weights (arXiv) (AISTATS 2019)
- Graph Embedding with Shifted Inner Product Similarity and Its Improved Approximation Capability (arXiv) (AISTATS 2019) (software)
- An information criterion for auxiliary variable selection in incomplete data analysis (Entropy 2019)
- Selective Inference for Testing Trees and Edges in Phylogenetics (arXiv 2019) (Frontiers in Ecology and Evolution) (software)
- Word-like character n-gram embedding (W-NUT 2018)
- Transitivity vs Preferential Attachment: Determining the Driving Force Behind the Evolution of Scientific Co-Authorship Networks (ICCS 2018)
- On representation power of neural network-based graph embedding and beyond (arXiv) (ICML 2018 workshop Theoretical Foundations and Applications of Deep Generative Models)
- A probabilistic framework for multi-view feature learning with many-to-many associations via neural networks (arXiv) (ICML 2018)
- Selective inference for the problem of regions via multiscale bootstrap (arXiv 2018)
- Segmentation-Free Word Embedding for Unsegmented Languages (EMNLP 2017)
- Spectral Graph-Based Method of Multimodal Word Embedding (TextGraphs-11 2017)
- An information criterion for model selection with missing data via complete-data divergence (Annals of the Institute of Statistical Mathematics 2018)
- Joint estimation of preferential attachment and node fitness in growing complex networks (Scientific Reports 2016)
- Image and tag retrieval by leveraging image-group links with multi-domain graph embedding (ICIP 2016)
- Cross-Lingual Word Representations via Spectral Graph Embeddings (ACL 2016) (CL-Eigenwords Website)
- Cross-validation of matching correlation analysis by resampling matching weights (Neural Networks 2016) (arXiv)
- PAFit: A Statistical Method for Measuring Preferential Attachment in Temporal Complex Networks (PLOS ONE 2015)
- Higher-order accuracy of multiscale-double bootstrap for testing regions (Journal of Multivariate Analysis 2014)
- Measuring preferential attachment in growing networks with missing-timelines using Markov chain Monte Carlo (Physica A 2012)
Old Technical Reports
- Improving predictive inference under covariate shift by weighting the log-likelihood function, ISM RM-712, 1998. This technical report is almost the same as the first paper of “covariate shift” adaptation (Shimodaira JSPI 2000), but its geometrical argument in Section 8 has been deleted when published.
- A graphical technique for model selection diagnosis, ISM RM-680, 1998. A part of this technical report is included in a paper (Shimodaira Comm. Stat. 2001) published later.