Tsuyoshi Okita

Tsuyoshi Okita

We propose SVAR-FM (Structural VAR with Flow Matching), a framework for time series causal discovery that treats a physics-based simulator as a mechanical realization of Pearl's do operator. Clamping a variable inside the simulator physically severs confounding paths, producing interventional data by construction. Conditional Flow Matching then learns the nonlinear interventional conditionals. Theoretically, we prove that the full structural VAR becomes identifiable under a coverage condition on the simulator-clampable variables, and derive an end-to-end error bound that decomposes into Monte Carlo, simulator fidelity, and Flow Matching terms. A sign-flip corollary predicts that when simulator accuracy falls below a threshold, the estimated causal effect reverses sign. Empirically, a benchmark across four scientific domains confirms that SVAR-FM recovers the correct causal sign where observational methods produce sign-reversed estimates due to confounding. A case study in ultrafast laser physics verifies the sign-flip prediction by physically varying the accuracy level of a first-principles quantum solver: the low-accuracy setting reverses the causal sign, while the high-accuracy setting recovers the correct direction (R-squared = 0.983, zero bias).

Kosuke Ukita, Tsuyoshi Okita

Diffusion models can learn rich representations during data generation, showing potential for Self-Supervised Learning (SSL), but they face a trade-off between generative quality and discriminative performance. Their iterative sampling also incurs substantial computational and energy costs, hindering industrial and edge AI applications. To address these issues, we propose the Flow Matching-based Foundation Model (FlowFM), which jointly trains a representation encoder and a conditional flow matching generator. This decoupled design achieves both high-fidelity generation and effective recognition. By using flow matching to learn a simpler velocity field, FlowFM accelerates and stabilizes training, improving its efficiency for representation learning. Experiments on wearable sensor data show FlowFM reduces training time by 50.4\% compared to a diffusion-based approach. On downstream tasks, FlowFM surpassed the state-of-the-art SSL method (SSL-Wearables) on all five datasets while achieving up to a 51.0x inference speedup and maintaining high generative quality. The implementation code is available at https://github.com/Okita-Laboratory/jointOptimizationFlowMatching.

Tsuyoshi Okita

Recent methods for improving LLM mathematical reasoning, whether through MCTS-based test-time search or causal graph-guided knowledge injection, cannot identify which concepts causally contribute to a correct answer, as the observed association may be spurious, driven by confounders such as problem difficulty. We propose CIKA (Causal Intervention for Knowledge Activation), a framework that uses the LLM itself as an interventional simulator: a prompt sets the concept state to ``mastered'' and the correctness change estimates the causal effect. We formalize this quantity as an Interventional Capability Probe (ICP), which diagnoses whether the LLM can use a given concept -- distinct from merely possessing knowledge. Because the intervention exogenously sets the concept state independently of problem difficulty, ICP separates confounding that observational methods cannot. On 67 screened problems, the ICP of the top-ranked concept (+0.219) is significantly larger than that of the negative control (+0.039; paired $t$-test, $p < 10^{-6}$, Cohen's $d = 0.86$), confirming that the probe discriminates causally relevant concepts from irrelevant ones. Analysis of 601 Omni-MATH problems further shows that solved problems have 6.1$\times$ higher ATE than unsolved ones (0.338 vs. 0.055), confirming that ICP is predictive of problem-solving success. With a 7B-parameter LLM whose weights are entirely frozen, CIKA achieves 69.7\% on the contamination-free Omni-MATH-Rule benchmark and 64.0\% overall, compared to 60.5\% for o1-mini, and 97.2\% on GSM8K, 46--50\% on AIME 2024--2026, and 46.2\% on MathArena. The Causal Knowledge Activation component contributes 33.8\% of correct answers on problems where the base model alone fails, demonstrating that the LLM already possessed but had not activated the requisite knowledge.

Tsuyoshi Okita

Extrapolation to out-of-distribution conditions is a fundamental challenge for physics-constrained deep generative models. Existing methods apply physical constraints as a single static regularization term uniformly across the generation process, and address neither the hierarchical structure of physical laws and the confounding variable problem. We propose the Deconfounded Hierarchical Gate (DHG), which serves as a diagnostic and control mechanism: it identifies when and how strongly temperature confounding contaminates each constraint level, so that hierarchical gates reflect intrinsic physical inconsistency rather than spurious temperature effects. DHG combines counterfactual estimation via the do-operator with backdoor adjustment to remove confounding, then applies Coarse-to-Fine physical constraints progressively. We report a counter-intuitive finding in pretraining: excluding the target-domain data from pretraining outperforms including it by 39% in extrapolation performance (RMSE 0.224 vs. 0.324). This occurs because FNO learns domain-agnostic physical patterns that transfer more effectively when the target domain is withheld. On a lithium-ion battery temperature extrapolation benchmark (trained at 24 degrees Celsius, evaluated at 4.0--43.0 degrees Celsius), our method achieves RMSE = 0.215, a 46% improvement over the unconstrained baseline (Pure CFM: 0.397).

Tsuyoshi Okita

Existing interventional causal discovery methods -- IGSP, DCDI, ENCO -- assume causal sufficiency (no latent confounders) and rely on virtual interventions in synthetic simulators. In AI-for-Science settings such as molecular design and materials science, latent confounders are ubiquitous and real interventions (e.g., physics-based simulations) require hours to days per data point. We propose CFM-SD (Causal Flow Matching with Simulation Data), which uses first-principles physical simulators as do-operators in Pearl's interventional calculus to simultaneously handle latent confounders and real interventional data. Theoretically, $d$-variable causal structure is identifiable with $O(d)$ single-variable interventions -- the minimum under physical realizability constraints. In Intrinsic Evaluation on synthetic data ($γ=0.2$--$0.8$), CFM-SD achieves average F1$=0.800$ vs. F1$=0.127$--$0.562$ for all baselines. In Extrinsic Evaluation on real scientific data, CFM-SD achieves 57--58\% bias reduction in molecular toxicity prediction and battery electrolyte optimization, demonstrating practical value beyond synthetic benchmarks.

Tsuyoshi Okita

Conventional time-series generation often ignores domain-specific physical constraints, limiting statistical and physical consistency. We propose a hierarchical framework that embeds the inherent hierarchy of physical laws-conservation, dynamics, boundary, and empirical relations-directly into deep generative models, introducing a new paradigm of physics-informed inductive bias. Our method combines Fourier Neural Operators (FNOs) for learning physical operators with Conditional Flow Matching (CFM) for probabilistic generation, integrated via time-dependent hierarchical constraints and FNO-guided corrections. Experiments on harmonic oscillators, human activity recognition, and lithium-ion battery degradation show 16.3% higher generation quality, 46% fewer physics violations, and 18.5% improved predictive accuracy over baselines.

Koki Matsuishi, Kosuke Ukita, Tsuyoshi Okita

In recent years, the widespread adoption of wearable devices has highlighted the growing importance of behavior analysis using IMU. While applications span diverse fields such as healthcare and robotics, recent studies have increasingly focused on multimodal analysis, in addition to unimodal analysis. Several studies have proposed multimodal foundation models that incorporate first-person video and text data; however, these models still fall short in providing a detailed analysis of full-body human activity. To address this limitation, we propose Activity Understanding and Representations Alignment - Multimodal Foundation Model (AURA-MFM), a foundational model integrating four modalities: third-person video, motion capture, IMU, and text. By incorporating third-person video and motion capture data, the model enables a detailed and multidimensional understanding of human activity, which first-person perspectives alone fail to capture. Additionally, a Transformer-based IMU encoder is employed to enhance the model's overall performance. Experimental evaluations on retrieval and activity recognition tasks demonstrate that our model surpasses existing methods. Notably, in the zero-shot classification for action recognition, our method achieved significantly higher performance, with an F1-score of 0.6226 and an accuracy of 0.7320, whereas the existing method recorded an F1-score of 0.0747 and an accuracy of 0.1961.

Masaharu Kagiyama, Tsuyoshi Okita

This paper aims to develop an energy-efficient classifier for time-series data by introducing PatchEchoClassifier, a novel model that leverages a reservoir-based mechanism known as the Echo State Network (ESN). The model is designed for human activity recognition (HAR) using one-dimensional sensor signals and incorporates a tokenizer to extract patch-level representations. To train the model efficiently, we propose a knowledge distillation framework that transfers knowledge from a high-capacity MLP-Mixer teacher to the lightweight reservoir-based student model. Experimental evaluations on multiple HAR datasets demonstrate that our model achieves over 80 percent accuracy while significantly reducing computational cost. Notably, PatchEchoClassifier requires only about one-sixth of the floating point operations (FLOPS) compared to DeepConvLSTM, a widely used convolutional baseline. These results suggest that PatchEchoClassifier is a promising solution for real-time and energy-efficient human activity recognition in edge computing environments.

Haruki Kai, Tsuyoshi Okita

We developed a deep learning algorithm for human activity recognition using sensor signals as input. In this study, we built a pretrained language model based on the Transformer architecture, which is widely used in natural language processing. By leveraging this pretrained model, we aimed to improve performance on the downstream task of human activity recognition. While this task can be addressed using a vanilla Transformer, we propose an enhanced n-dimensional numerical processing Transformer that incorporates three key features: embedding n-dimensional numerical data through a linear layer, binning-based pre-processing, and a linear transformation in the output layer. We evaluated the effectiveness of our proposed model across five different datasets. Compared to the vanilla Transformer, our model demonstrated 10%-15% improvements in accuracy.

Kunpeng Zhao, Asahi Miyazaki, Tsuyoshi Okita

Human Activity Recognition (HAR) has recently witnessed advancements with Transformer-based models. Especially, ActionFormer shows us a new perspectives for HAR in the sense that this approach gives us additional outputs which detect the border of the activities as well as the activity labels. ActionFormer was originally proposed with its input as image/video. However, this was converted to with its input as sensor signals as well. We analyze this extensively in terms of deep learning architectures. Based on the report of high temporal dynamics which limits the model's ability to capture subtle changes effectively and of the interdependencies between the spatial and temporal features. We propose the modified ActionFormer which will decrease these defects for sensor signals. The key to our approach lies in accordance with the Sequence-and-Excitation strategy to minimize the increase in additional parameters and opt for the swish activation function to retain the information about direction in the negative range. Experiments on the WEAR dataset show that our method achieves substantial improvement of a 16.01\% in terms of average mAP for inertial data.

Gao Huayu, Huang Tengjiu, Ye Xiaolong et al.

AI-based motion capture is an emerging technology that offers a cost-effective alternative to traditional motion capture systems. However, current AI motion capture methods rely entirely on observed video sequences, similar to conventional motion capture. This means that all human actions must be predefined, and movements outside the observed sequences are not possible. To address this limitation, we aim to apply AI motion capture to virtual humans, where flexible actions beyond the observed sequences are required. We assume that while many action fragments exist in the training data, the transitions between them may be missing. To bridge these gaps, we propose a diffusion-model-based action completion technique that generates complementary human motion sequences, ensuring smooth and continuous movements. By introducing a gate module and a position-time embedding module, our approach achieves competitive results on the Human3.6M dataset. Our experimental results show that (1) MDC-Net outperforms existing methods in ADE, FDE, and MMADE but is slightly less accurate in MMFDE, (2) MDC-Net has a smaller model size (16.84M) compared to HumanMAC (28.40M), and (3) MDC-Net generates more natural and coherent motion sequences. Additionally, we propose a method for extracting sensor data, including acceleration and angular velocity, from human motion sequences.

Koki Matsuishi, Tsuyoshi Okita

In deep multi-instance learning, the number of applicable instances depends on the data set. In histopathology images, deep learning multi-instance learners usually assume there are hundreds to thousands instances in a bag. However, when the number of instances in a bag increases to 256 in brain hematoma CT, learning becomes extremely difficult. In this paper, we address this drawback. To overcome this problem, we propose using a pre-trained model with self-supervised learning for the multi-instance learner as a downstream task. With this method, even when the original target task suffers from the spurious correlation problem, we show improvements of 5% to 13% in accuracy and 40% to 55% in the F1 measure for the hypodensity marker classification of brain hematoma CT.

Asahi Miyazaki, Tsuyoshi Okita

Masked image modeling is one of the most poplular objectives of training. Recently, the SparK model has been proposed with superior performance among self-supervised learning models. This paper proposes a new mask pattern for this SparK model, proposing it as the Mesh Mask-ed SparK model. We report the effect of the mask pattern used for image masking in pre-training on performance.

Kosuke Ukita, Ye Xiaolong, Tsuyoshi Okita

In this paper, we propose a diffusion model that integrates a representation-conditioning mechanism, where the representations derived from a Vision Transformer (ViT) are used to condition the internal process of a Transformer-based diffusion model. This approach enables representation-conditioned data generation, addressing the challenge of requiring large-scale labeled datasets by leveraging self-supervised learning on unlabeled data. We evaluate our method through a zero-shot classification task for hematoma detection in brain imaging. Compared to the strong contrastive learning baseline, DINOv2, our method achieves a notable improvement of +6.15% in accuracy and +13.60% in F1-score, demonstrating its effectiveness in image classification.

Kodai Hirata, Tsuyoshi Okita

This paper proposes a method MTL-Swin-Unet which is multi-task learning using transformers for classification and semantic segmentation. For spurious-correlation problems, this method allows us to enhance the image representation with two other image representations: representation obtained by semantic segmentation and representation obtained by image reconstruction. In our experiments, the proposed method outperformed in F-value measure than other classifiers when the test data included slices from the same patient (no covariate shift). Similarly, when the test data did not include slices from the same patient (covariate shift setting), the proposed method outperformed in AUC measure.

Keita Nonaka, Kazutaka Yamanouchi, Tomohiro I et al.

In this study, we propose a simple and effective preprocessing method for subword segmentation based on a data compression algorithm. Compression-based subword segmentation has recently attracted significant attention as a preprocessing method for training data in Neural Machine Translation. Among them, BPE/BPE-dropout is one of the fastest and most effective method compared to conventional approaches. However, compression-based approach has a drawback in that generating multiple segmentations is difficult due to the determinism. To overcome this difficulty, we focus on a probabilistic string algorithm, called locally-consistent parsing (LCP), that has been applied to achieve optimum compression. Employing the probabilistic mechanism of LCP, we propose LCP-dropout for multiple subword segmentation that improves BPE/BPE-dropout, and show that it outperforms various baselines in learning from especially small training data.

Hokuto Hirano, Tsuyoshi Okita

Cerebral hematoma grows rapidly in 6-24 hours and misprediction of the growth can be fatal if it is not operated by a brain surgeon. There are two types of cerebral hematomas: one that grows rapidly and the other that does not grow rapidly. We are developing the technique of artificial intelligence to determine whether the CT image includes the cerebral hematoma which leads to the rapid growth. This problem has various difficulties: the few positive cases in this classification problem of cerebral hematoma and the targeted hematoma has deformable object. Other difficulties include the imbalance classification, the covariate shift, the small data, and the spurious correlation problems. It is difficult with the plain CNN classification such as VGG. This paper proposes the joint learning of semantic segmentation and classification and evaluate the performance of this.

Tsuyoshi Okita, Hirotaka Hachiya, Sozo Inoue et al.

The understanding of sensor data has been greatly improved by advanced deep learning methods with big data. However, available sensor data in the real world are still limited, which is called the opportunistic sensor problem. This paper proposes a new variant of neural machine translation seq2seq to deal with continuous signal waves by introducing the window-based (inverse-) representation to adaptively represent partial shapes of waves and the iterative back-translation model for high-dimensional data. Experimental results are shown for two real-life data: earthquake and activity translation. The performance improvements of one-dimensional data was about 46% in test loss and that of high-dimensional data was about 1625% in perplexity with regard to the original seq2seq.

Tsuyoshi Okita

A neural probabilistic language model (NPLM) provides an idea to achieve the better perplexity than n-gram language model and their smoothed language models. This paper investigates application area in bilingual NLP, specifically Statistical Machine Translation (SMT). We focus on the perspectives that NPLM has potential to open the possibility to complement potentially `huge' monolingual resources into the `resource-constraint' bilingual resources. We introduce an ngram-HMM language model as NPLM using the non-parametric Bayesian construction. In order to facilitate the application to various tasks, we propose the joint space model of ngram-HMM language model. We show an experiment of system combination in the area of SMT. One discovery was that our treatment of noise improved the results 0.20 BLEU points if NPLM is trained in relatively small corpus, in our case 500,000 sentence pairs, which is often the case due to the long training time of NPLM.