A new approach is developed for estimating bivariate SVAR systems with Gaussian distribution shocks. Each sequence of independent shocks has respectively its own unknown variance parameter and a structural matrix also leaves their non-diagonal elements unknown. So, there are four unknown parameters, while the symmetric covariance matrix of the residuals in reduced VAR gives us three components. Estimation procedures consist of two stage random perturbations. First, a new variable with independent Gaussian distributions is, as a trial, added to the original 2-d system. By setting a structural matrix with three unknown parameters and taking three variance parameters into consideration, the number of unknown parameters becomes six, and we can solve this 3d SVAR system whose covariance matrix gives us six components. Among N trials of new variables, we select the best one so that two black lines of 2d and 3d systems are as close as possible. Secondly, the 3d system with the selected Gaussian sequence is perturbed by the other Gaussian sequences as exogenous variables. Then, in simulation experiments, we can find out the estimation criterion such that the p-value of estimated parameter in 3d system is minimized at the true parameter value in 2d system.

This study proposes a novel approach to improve forecast accuracy and uncertainty quantification in time series forecasting through the synthesis of Machine Learning (ML) models with Dynamic Bayesian Predictive Synthesis (DBPS). We synthesize predictive distributions from three ML models—LightGBM, LSTM, and Prophet—through DBPS, notably employing Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models to estimate uncertainties for ML forecasts. Empirical analysis using meteorological data from Delhi, India, demonstrates that our proposed method outperforms individual ML models and conventional ensemble methods, achieving improvements of up to 25.2% in RMSE and 25.4% in MAE. Furthermore, evaluating predictive distributions using Log Predictive Density Ratio (LPDR) reveals superior performance to other models, indicating a more accurate capture of observational uncertainty. Analysis of DBPS weights reveals that the offsetting effect of prediction biases is one of the factors contributing to improved forecast accuracy. These findings suggest that the synthesis of ML models and DBPS enables more reliable forecasting and improved uncertainty quantification.

In this paper, we focus on discrete-time random walks (DTRWs) on the path graph with reflecting walls. Also we consider the corresponding continuous-time random walks (CTRWs), discrete-time quantum walks (DTQWs) and continuous-time quantum walks (CTQWs). We give formulas for the return probabilities of these four walks (DTRWs, CTRWs, DTQWs, CTQWs) at the boundary point of the path graph and discuss their relationships. Furthermore, by considering the Ehrenfest model, we derive the scaling limit theorems for the return probabilities of both the CTRW and the CTQW.

The stochastic approximation is an iterative algorithm used to solve an unknown equation via noise corrupted observation of its residual. This paper develops its stopping rule under a martingale difference noise. This rule is a sufficient number of iterations to theoretically guarantee the distance between a current solution candidate and its solution. Its effectiveness is demonstrated through a numerical example.

In this paper, it is assumed that proceeds of cracks occur due to the degradation of materials, and the degradation is estimated by observations of the crack proceeds. Although numerical approaches had been proposed to evaluate the reliability based on degradation models, a hierarchical Bayesian method for the reliability evaluation of the degradation phenomenon is proposed. Moreover, comparison methodologies for the models based on the information criteria are proposed.

This study focuses on the development of a simulation model for an electric servo injection molding machine. Plastic products have become indispensable in a wide variety of fields due to their light weight, durability, formability, and insulation properties. Injection molding, an important process in its production, enables highly efficient mass production and high-precision shape molding. However, it faces challenges such as dependence on skilled engineers, variation in molding quality due to trial and error in the prototype stage, and waste of resources. To address these issues, this study employed a gray-box modeling approach that utilizes actual equipment data. In this study, we focused on injection equipment and built a base model to reproduce the molding process using Simscape (MATLAB). Parameters were then adjusted through Bayesian optimization to improve simulation accuracy. Initial results confirmed the validity of the model, and subsequent adjustments have significantly improved the correlation between simulation output and actual equipment data. The results of this research show the potential to contribute to efficient optimization of molding conditions, reduction of waste in the prototype stage, and improvement of operational efficiency.

The existence and stability of equilibrium points in nonlinear systems are often challenged by parameter variations, which can compromise traditional stability analysis methods. This paper addresses the parametric stabilization of uncertain nonlinear systems characterized by parameter-dependent coefficients and a nonlinear function satisfying a sector condition. A linear state feedback control law is proposed, with a parameter dependent LMI condition ensuring the existence and quadratic stability of an equilibrium point in the closed loop system. To handle parameter uncertainties, a sequential randomized algorithm is employed, providing a probabilistic solution that achieves parametric quadratic stabilization with high confidence.

The stochastic shortest path problem (SSP) is a standard model for sequential decision making under uncertain environments. In addition, a method was proposed to explicitly handle failure scenarios by introducing dead-ends[4]. On the other hand, due to modeling errors with the real environment, the policy constructed for the model may not achieve the desired performance. In this paper, we derive a policy that works well by keeping the failure probability within a desired range even when the model has uncertainty. This problem is formulated as a partially observable stochastic dynamic game (POSDG) by introducing a label that indicate either the objective function or the constraints. Numerical examples are used to demonstrate the effectiveness of the obtained optimal policy and the property that action selection depends not only on the initial state and current state but also on the history.

In this study, we predict the flowering date of Somei-Yoshino across Japan using a one-dimensional convolutional neural network (1D-CNN) model with daily mean temperature as input. By combining this predictive model with MC dropout, we obtain prediction distributions. Furthermore, applying SHAP (SHapley Additive exPlanations) to the model with MC dropout allows us to obtain SHAP values for each feature of individual data samples, enabling an analysis of their distribution. By consistently identifying characteristics of SHAP value distributions, we can assess the reliability of features in models with inherent uncertainties. We expect this approach to provide new insights that traditional point estimation methods for predicting flowering dates cannot offer. In this study, we first analyze the consistency between the base model’s SHAP values and domain knowledge about cherry blossoms to verify the model’s validity. Next, we select specific locations and years to eliminate variability by location and year to compare temperatures and SHAP values, thereby deepening our understanding of how regional climate conditions affect flowering dates. Subsequently, we apply MC dropout to this base model to obtain SHAP value distributions and conduct a meticulous analysis of their shapes. Finally, based on these analyses, we investigate the relationship between the variance of prediction distributions and SHAP value distributions and quantify the reliability of feature influences with uncertainties on flowering dates. This comprehensive analysis will improve model interpretability and prediction accuracy.

In recent years, research on Brain-Computer Interface (BCI) technology using electroencephalography (EEG) based on motor imagery has gained attention as a potential means of supporting the lives of individuals with physical disabilities. However, conventional EEG devices require lengthy setup times, placing a significant burden on users. This study aims to classify motor intentions by processing data obtained from left-hand and right-hand motor imagery experiments using a simplified EEG device. Previous studies have attempted feature extraction using Event-Related Synchronization/Desynchronization (ERS/ERD), but only partial features were identified, resulting in low classification accuracy. On the other hand, using Common Spatial Patterns (CSP) has been reported to improve accuracy. To validate the effectiveness of CSP, we conducted experiments using the dataset provided by the Berlin BCI Group. The results demonstrated that CSP successfully highlighted differences in features between lefthand and right-hand motor imagery. Although CSP also improved accuracy compared to ERS/ERD in the dataset from our laboratory, further improvements are still necessary

Vehicle traffic congestion is a common problem in many cities around the world. Many congestion reduction strategies have been proposed in the past, ranging from roadway infrastructure extension to transportation demand management. A highly effective way to reduce traffic congestion is an adaptive traffic signal control (ATSC) system. In this study, a novel ATSC system with decentralized reinforcement learning is proposed. Our ATSC system is developed under the non-Markovian Decision Process (NMDP) framework, in which each intersection corresponds to each state in Q-learning, and it works to alleviate effectively traffic congestion with the similar schemes of the Q-learning. As result, it is shown that our proposed system can mitigate effectively the dimension of the action-state space.

The task of exchanging surgical instruments between the surgeon and scrub nurse has been targeted for automation using a scrub nurse robot (SNR) [1,2,3,4,5]. To make the SNR system practical, real-time recognition of the surgeon's intentions is crucial. To achieve this, a classification model called Temporal Pose Feature Convolutional Neural Network (TPF CNN) [6,7] has been developed to recognize surgical procedures based on body movements extracted from skeleton data. In this paper, we propose a new model and report on two modifications aimed at the practical application of the system. Specifically, we introduce the TPF Attention Network (TPF AN), a novel model incorporating a self-attention mechanism [12] that is better suited for processing sequential data. The first modification involves adding a class to recognize when the surgeon is not performing a surgical action. This addition enables continuous inference throughout the entire surgical procedure, from start to finish. The second modification involves evaluating the system using the Leave-Person-Out (LPO) framework. Unlike the Leave-Videos-Out (LVO) framework used in prior research, which included at least one video from each surgeon in the training set, the LPO framework uses all videos from one surgeon for testing while the remaining surgeons' videos are used for training. This ensures a robust evaluation of the model's performance when encountering surgeons not included in the training data. Experimental results show that the proposed model improvements led to a 2.0-point increase in accuracy compared to conventional methods.

In object detection tasks, users generally can adjust the confidence score threshold to control the balance between false positives and missed detections. Bounding boxes with confidence scores exceeding this threshold are included in the final detection results. However, this creates a trade-off: raising the threshold reduces false positives but increases missed detections, while lowering the threshold has the opposite effect. This trade-off poses a challenge, as threshold adjustment alone is insufficient to effectively address both issues simultaneously. To overcome this limitation, this study proposes a hybrid model called GBDT-YOLOX, which combines a Gradient Boosting Decision Tree (GBDT) model with the detection results produced by YOLOX, one of the state-of-the-art object detection model. The GBDT model refines the raw detection outputs by leveraging additional contextual and feature-level information, enabling more informed decisions and reducing reliance on a single threshold. Experimental results on the COCO dataset demonstrate that GBDTYOLOX achieves 1.5% and 3.0% improvements in Average Precision (AP) and Average Recall (AR), respectively. These results indicate that the proposed method effectively reduce both false positives and missed detections. Additionally, to evaluate its performance in a more challenging domain-specific scenario, experiments were conducted on the RailSem19 dataset, which focuses on railway environments with unique detection challenges. The results showed an approximate 3% improvement in AP, further validating the effectiveness of GBDT-YOLOX in reducing detection errors in complex and specialized applications.

In application fields requiring high safety standards, such as autonomous driving and railway systems, the accurate detection of critical objects is essential for system reliability. This study focuses on improving the detection accuracy of railway switches, key components that control train direction, using the RailSem19 dataset. Switches are visually similar to other structures and backgrounds in railway environments, making their detection inherently challenging. We propose a two-stage method that combines YOLOX, a state-of-the-art object detection model, with Efficient-NetB2 for fine-grained classification. YOLOX detects candidate regions for switches, which are refined by Efficient-NetB2 to classify them into three sub-categories (switchright, switch-left, and switch-unknown) or background. Experiments on the RailSem19 dataset showed that the proposed methods improved detection accuracy (AP0.5:0.95) by approximately 5 points for switch-right and switch-left, and by 0.6 points for switch-unknown, compared to YOLOX alone. Additionally, applying the same mechanism to YOLOv10, the latest model of YOLO series, demonstrated similar improvements.

This study investigates the neural mechanisms underlying the recognition of English sounds “L” and “R,” which are challenging for Japanese learners due to the absence of equivalent phonemes in their native language. Using eventrelated potentials (ERPs), we analyze neural responses to visual (letter) and auditory (sound) stimulus. Participants are presented with the English words “right” and “light” in four stimulus combinations. They press a key only when they feel stimulus are matched. Across 500 randomized trials, ERPs reveal faster and more distinct responses to visual stimulus compared to auditory ones. Participants exhibit greater difficulty in processing mismatched stimulus involving “L” and “R,” highlighting the influence of phonetic unfamiliarity on neural activity. These findings highlight the role of attention in sensory processing during language learning. Visual support enhances auditory discrimination, offering insights into pedagogical strategies and tools to improve pronunciation and phonetic recognition.

Improving the pronouncing skills of English vowels is difficult for those whose first language is Japanese. Due to the differences in the variation of vowels between Japanese and English, some vowels are difficult for native Japanese speakers. Articulating vowels are a consecutive movement of lips and tongue. Therefore, the usage of speakers’ facial muscles affects the accuracy of their pronunciation. We have done an experiment to find out whether the photograph of instructor’s mouth is effective for pronunciation training.

To manage daily sleep, continuous monitoring of vital signs such as heart rates and respiration rates during sleep are required. In this study, we are studying a heart rate extraction method using an infrared camera to extract biological information from sleeping subjects without contact and restraint. We have developed a method to extract information related to heart rate by applying independent component analysis (ICA) to signals extracted from multiple subregions of the face, and confirmed its effectiveness．In this paper, we propose an improved method to improve the accuracy of extraction by dividing the acquired time-series data by short time periods and applying ICA repeatedly. Furthermore, a Long Short-Term Memory (LSTM) model is introduced to determine which of the extracted signals contains heart rate information. In an evaluation experiment, we tried to extract heart rate information through a nap experiment of about 60 minutes. As a result of analysis in the proposed method, heart rate information was successfully extracted.

Falls in daily life can be a major cause of injury for many people. Preventing falls is essential for extending healthy lifespan. Various exercises aimed at fall prevention have been proposed, but their effectiveness is often evaluated based on empirical knowledge, and quantitative assessments of their effectiveness have not been sufficiently developed. This paper proposes a method for quantitatively evaluating the effects of fall prevention exercises by focusing on muscle activity and the movement of the center of mass, with the goal of assessing the effectiveness of the exercises in reducing fall risk. To evaluate the effects of fall prevention exercises, the Mahalanobis distance relative to a quiet standing posture was calculated based on surface electromyography (sEMG) data of muscles involved in postural control and center of mass data, and an analysis was conducted. In addition, contribution rates for each variable were calculated to identify the muscles and center of mass movements contributing to postural control. The analyzed movements included Judo Kenko Exercises, a fall prevention exercise, as well as Radio Exercises, an exercise aimed at promoting health, and walking, a daily activity that requires dynamic balance. The results suggested that Judo Kenko Exercises could potentially train not only the gluteus maximus, which is commonly necessary for postural control in standing, but also muscles used for rapid fall-avoidance movements. Furthermore, it was suggested that Judo Kenko Exercises might be a safer training method with a lower fall risk compared to walking.

This study investigated the impact of electrode aperture size on cortical surface potential recordings, focusing on electrocorticography (ECoG) signals and addressing concerns about spatial aliasing and signal amplitude variations in high-density ECoG recordings. Two ECoG electrode arrays with different aperture dimensions (200 µm and 460 µm) were fabricated and tested in the rat visual cortex. Visually evoked potentials (VEPs) were recorded and analyzed for temporal and spatial variations. The results showed that 460 µm-aperture electrodes detected higher amplitude responses in certain regions compared to 200 µm -aperture electrodes, especially for spatially localized VEP peaks. This was attributed to the larger aperture size of the 460 µm electrodes, allowing the detection of VEP components missed by the smaller apertures. These findings align with theoretical considerations, indicating that the electrode aperture size significantly affects the measured signals in ECoG recordings.

Positioning methods that do not rely on GNSS are urgently needed to address vulnerabilities. In this respect, using LEO constellation for navigation has attracted much attention in recent years. In this study, we focus on the ORBCOMM satellite and aim to obtain the Doppler frequency using signal processing. We show that the frequency offset — the difference between transmitted and received signal frequencies — can be estimated even in an environment where the observer does not have orbit information of the satellite by using the modulation scheme of the satellite.

This paper proposes a real-time GNSS spoofing detection method using a simplified signal anomaly classification approach. It utilizes readily available parameters from observation data, focusing on signal strength and satellite elevation information. The method employs statistical measures and the IQR method to evaluate signal characteristics, establishing a classification flow from "Definite Spoofing" to "Good Situation". Spoofing experiments, including non-coherent and coherent attacks, were conducted to evaluate the method's performance. This approach aims to provide a simple, effective, and widely implementable method for enhancing GNSS security across various applications.