Reservoir computing using nonlinear vortex dynamics in type-II superconductors enables low-power time series processing, though accuracy has been limited. To examine performance factors, 2D time-dependent Ginzburg-Landau simulations were conducted with varied pin density, pinning strength, and temperature. Pinning was controlled via local α values, and temperature via bulk α. Input current and output electric field formed the input-output pair, evaluated by NARMA2 accuracy and memory capacity. Results showed optimal performance at moderate pinning and low temperatures. Irregular responses at low temperature were linked to enhanced vortex-pin interactions, offering design insights for high-precision superconducting reservoir hardware.

Reservoir computing, a machine learning paradigm derived from recurrent neural networks, is efficient for complex time-series prediction tasks. Its advantages, including reduced computational requirements for weights, real-time prediction capability, and the use of physical nonlinear dynamical systems, make it well-suited for edge computing. However, acquiring the transient-state responses of reservoirs using multiple physical nodes requires substantial memory storage and numerous recording channels, which hinders dense integration and usage in memory-constrained devices. In this work, we present a frequency domain multiplexing approach to effectively utilize the collective dynamics of a coupled spin Hall oscillator array using a single output node. We investigate the collective behavior of the coupled oscillators through micromagnetic simulations and analyze reservoir states in both the time domain and Fourier space using Mackey-Glass inputs. Our results indicate that spectral analysis in the Fourier domain enhances reservoir performance, offering a promising strategy for data processing in physically constrained reservoir computing systems.

Modeling how signal transduction pathways shape stimulus-evoked membrane-voltage dynamics is essential for linking molecular perturbations to computation. Existing approaches are either costly biophysical models or black-box systems, obscuring pathway roles and preventing computational knockouts. We propose a modular reservoir architecture, the Sequential Multi-Output Echo State Network (SMO-ESN), which partitions the reservoir into serial modules and applies structured dropout to mask module outputs. Trained on experimental voltage recordings from C. elegans AWA neurons under step odor stimulation in wild-type and an egl-19 null mutant, SMO-ESN achieves lower NRMSE than a baseline ESN and reproduces mutant-like responses, highlighting interpretability.

Fabricated in-materio reservoir computing devices have demonstrated ultra-low power consumption for unconventional computing tasks across varieties of material systems. Simulating their internal networks is desired to analyze them, but challenging due to physical variability. While echo state networks (ESNs) provide a relevant framework, conventional ESN simulations require a long computational time at large scales and fail to replicate physical device outputs. We propose a novel ESN framework incorporating structural constraints inspired by physical in-materio devices, featuring a unidirectional cluster-based topology. Our approach demonstrates high consistency, faster computation, and promising benchmark performance compared to conventional ESNs.

We propose novel methods for enhancing the memory capacity of reservoir computing (RC) solely by modifying the network configuration without changing the RC layer. Delay, Passthrough, and Parallel methods are introduced and applied to the echo state network (ESN) and chaotic Boltzmann machine (CBM-)RC. Evaluations using the NARMA task and information processing capacity (IPC) showed that these methods significantly improve memory capacity while enabling control over the memory-nonlinearity trade-off. The Delay-Passthrough combination yielded the best performance across models, particularly benefiting CBM-RC, which is advantageous for analog VLSI implementations where internal modification is constrained.

Sign language recognition (SLR) facilitates communication between deaf and hearing communities. Deep learning based SLR models are commonly used but require extensive computational resources, making them unsuitable for deployment on edge devices. To address these limitations, we propose a lightweight SLR system that combines parallel bidirectional reservoir computing (PBRC) with MediaPipe. MediaPipe enables real-time hand tracking and precise extraction of hand joint coordinates, which serve as input features for the PBRC architecture. The proposed PBRC architecture consists of two echo state network (ESN) based bidirectional reservoir computing (BRC) modules arranged in parallel to capture temporal dependencies, thereby creating a rich feature representation for classification. We trained our PBRC-based SLR system on the Word-Level American Sign Language (WLASL) video dataset, achieving top-1, top-5, and top-10 accuracies of 60.85%, 85.86%, and 91.74%, respectively. Training time was significantly reduced to 18.67 seconds due to the intrinsic properties of reservoir computing, compared to over 55 minutes for deep learning based methods such as Bi-GRU. This approach offers a lightweight, cost-effective solution for real-time SLR on edge devices.

We enhance bidirectional 2-dimensional (2D) reservoir computing (BiRC2D) by incorporating hierarchical pooling operations for computer vision tasks on edge devices. BiRC2D effectively captures local spatial dependencies in image data while being based on reservoir computing. However, BiRC2D lacks downsampling capabilities, limiting the ability to capture multi-scale image structures. To address this limitation, we introduce a hierarchical extension by alternately stacking BiRC2D blocks and 2D pooling layers. This enhancement enables progressive spatial feature abstraction while preserving the low-parameter, training-free advantages of reservoir computing. Anomaly detection experiments on the MVTec AD dataset demonstrate that feature embedding-based methods using our proposed architecture achieve competitive performance while reducing the parameter count by 97-99% compared to those using ResNet-50. Our proposed architecture operates solely through random spatial dynamics, offering efficient and scalable anomaly detection. These properties make it particularly well-suited for energy-constrained, real-time industrial inspection systems.

Home-service robots are expected to enhance the quality of daily life. Two types of knowledge, commonsense and environment-specific knowledge, are required for home-service robots. Current robots can handle audio-visual-based common knowledge, task planning, and action generation. Additionally, several systems that handle environment-specific knowledge have been developed. Previously proposed brain-inspired models can integrate visual and location information to represent episodes; however, it is difficult to obtain novel environment-specific knowledge. In this study, we propose a system that efficiently acquires novel environment-specific knowledge by combining environment-specific knowledge stored in a brain-inspired memory system with commonsense knowledge inferred by large language models (LLMs). We verified the performance of common-sense retrieval from LLMs and evaluated the effectiveness of combining environment-specific knowledge and commonsense knowledge in the home environment.

Dynamical Network Biomarkers (DNB) theory has recently emerged as a promising framework for the ultra-early detection of diseases, particularly when dealing with high-dimensional low-sample-size (HDLSS) data. Once such early warning signs are identified, timely intervention becomes essential to prevent the disease from progressing into irreversible states. From the perspective of control theory, such intervention aims to improve the system stability margin to avoid critical transitions, a process known as re-stabilization. Successful re-stabilization requires knowledge of the system parameters. However, the HDLSS nature of biological datasets poses significant challenges for precise system parameter identification. To address this issue, this study explores the application of the extended Kalman filter (EKF) and proposes a novel dual-loop EKF approach. In the inner-loop iteration, we simultaneously estimate both the system states and unknown parameters by augmenting them into a unified model and employing EKF with first-order linearization. Meanwhile, the outer loop iteratively refines these parameter estimates by reusing historical measurement data, eliminating the need for additional data collection. Numerical simulations on low- and high-dimensional systems demonstrate that the proposed dual-loop EKF method significantly improves parameter estimation accuracy compared to the traditional EKF method, highlighting its potential applicability in complex biological contexts.

This paper addresses a covariance steering problem for linear dynamical systems. A key feature of the problem is that the system model is unknown, and only a few snapshot data points on state variations are available. To address this challenge, the problem is decomposed into two parts: data-driven modeling based on snapshot data and a covariance steering problem toward the desired distribution using the estimated model. For the data-driven modeling problem, we propose a solution method based on the Schrödinger bridge formulation, as studied in prior work. For the model-based covariance steering problem, we develop a gradient-based algorithm in which the gradient is computed using two solutions of Lyapunov equations. Finally, we present a numerical simulation to demonstrate the effectiveness of the proposed approach.

To extend healthy life expectancy in an aging society, it is crucial to prevent various diseases at pre-disease states. Although dynamical network biomarker theory has been developed for pre-disease detection, mathematical frameworks for pre-disease treatment have not been well established. Here I propose a control theory-based approach for pre-disease treatment, named Markov chain sparse control (MCSC), where time evolution of a probability distribution on a Markov chain is described as a discrete-time linear system. By designing a sparse controller, a few candidate states for intervention are identified. The validity of MCSC is demonstrated using numerical simulations and real-data analysis.

Dynamical network biomarker (DNB) theory has emerged as a powerful framework for detecting early warning signals of pre-disease states. Based on our previous work demonstrating its utility in adipose tissue of metabolic syndrome model mice, we conducted a comprehensive DNB analysis using RNA sequencing data across 13 organs and 14 to 16 time points in high-fat diet-fed mice. Our findings revealed organ-specific variation in the timing of early warning signals, suggesting heterogeneous inter-organ dynamics during the pre-disease state of metabolic syndrome. These results highlight the potential of DNB theory for elucidating systemic early-stage pathophysiology in complex metabolic disorders.

Energy landscape analysis (ELA) offers a robust framework for examining state transitions in high-dimensional data, such as neuronal activity or microbial dynamics. The structure of the transition network from ELA depends heavily on the selected features. This study introduces a feature selection method using a scoring metric that measures orthogonality among binary representations of ELA-identified states. We applied this method to Specific Health Checkup data from Toyama Prefecture, Japan, and confirmed its effectiveness. Our results show that selecting features with this approach allows the ELA-based network to better capture major transitions in health status.

We propose a method for detecting abnormal cardiac rhythms—such as those associated with atrial fibrillation (AF)—from pulse waves measured by an automated blood pressure monitor. The method is based on reservoir computing and detects anomalies by evaluating the prediction error of a model trained solely on normal (sinus rhythm, SR) data. Hyperparameters were tuned using a validation set containing both SR and AF samples. Our method achieved a sensitivity of 100.0%, specificity of 98.6%, and accuracy of 99.2% on an independent test set, outperforming previously reported results. Our results suggest its potential as a cost-effective and accurate embedded screening device for identifying abnormal cardiac rhythms in daily health monitoring.

A photoplethysmogram (PPG) is an optical biomedical signal measuring changes in the amount of light reflected or transmitted through the skin. In recent years, imaging PPG(iPPG), which can measure pulse waves through video recordings by conventional cameras in a similar way as conventional PPG devices, has been attracting attention. In this study, we investigated the effect of measurement distance on the dynamic properties of iPPG and verified whether the dynamic properties remain. Using recurrence quantification analysis and the Wayland test on data from twelve healthy subjects, we found that iPPG maintains deterministic properties within 0.0 to 0.6 cm measurement distance. Also, while some measures of periodicity and regularity showed distance-dependent changes, we confirmed that the fundamental chaotic dynamics were maintained. These findings show that iPPG is feasible for non-contact health monitoring applications with advanced analysis, thus contributing to the improvement of biomedical sensing technology.

Time-series forecasting is directly linked to the operational optimization of social infrastructure, including electric power, transportation, weather, and infectious disease control. Various physical, statistical, and deep learning models have been developed, but all face a trade-off among interpretability, expressive power, and computational load. The Transformer, introduced in 2017, was initially hailed as a definitive solution for time-series forecasting due to its ability to capture long-term dependencies in parallel. However, a report at AAAI 2023 revealed cases where it failed to outperform even simple linear regression, casting significant doubt on its superiority. Since then, various improvements have been proposed, such as patching, channel independence, frequency decomposition, and Decoder-only foundation models. Nevertheless, as of 2025, fundamental challenges remain, including the sharpness of the loss function and the interpretability of attention. As a result, non-attention-based architectures, exemplified by GraphCast, are returning to the mainstream. This survey systematically organizes major research published from early 2023, when skepticism toward time-series Transformers grew, to the end of 2024. It formulates the basic structure of the Transformer and the task of time-series forecasting, and introduces the research progress of time-series forecasting methods using Transformers. We hope this paper will serve as a valuable reference not only for Transformer researchers but also for the broader research community engaged in the study of complex and dynamical systems.

The capacitated vehicle routing problem (CVRP) is one of the NP-hard combinatorial optimization problems, and efficient construction of a high-quality solution is a major challenge. In this paper, we propose a metaheuristic method based on chaotic dynamics to improve the efficiency of two local search operations: the CROSS-Exchange and the 2-opt methods. Unlike traditional methods that usually rely on randomization or fixed rules, this method dynamically controls the selection and execution of local search operations using chaotic neurodynamics. The experimental results show that this method can effectively improve the quality of the solution and algorithm performance when solving CVRP, and also provides a new solution framework for logistics optimization.

Over several decades, various methods and devices have been used to analyze circadian rhythms and sleep quality. However, conventional approaches struggle to capture long-term biological signal fluctuations. This study explores circadian rhythms and sleep quality in real-world settings using Fitbit and multiscale fuzzy entropy analysis. Findings indicate that heart rate variability (HRV) over a long timescale (approximately 3000 s) is associated with sleep efficiency. Although the underlying neural mechanisms remain unclear, understanding this pattern is crucial. Future research integrating long-term multimodal measurements with devices like Fitbit and portable electroencephalography devices could offer deeper insights into the sleep-HRV relationship.

Excitation phenomena in nonlinear circuits are realized using light-sensitive multivibrators that alternate between quiescent and oscillatory modes under external illumination. Circuit dynamics are governed by the continuous-time evolution of an analog internal state (capacitor voltage). Simultaneously, the output switches discretely between high and low levels, forming a hybrid system with mixed continuous-discrete behavior. Crucially, the switching threshold is not fixed but modulated by both the output state and the incident light, enabling tunable excitability. A class of such circuits was designed and theoretically analyzed, and their spike-like transient responses were quantitatively verified experimentally. The proposed configuration offers a compact platform for realizing spatially extended excitable media with discrete outputs that continuously evolve.

In regions with heavy snowfall, the living environment is becoming a serious problem due to heavy snow accumulation. A Road heating is an electrical device which promotes snow melting by burying a heating cable as a thermal source underground in such regions. When integrating the road heating into power distribution systems, we need to optimize the flow of electric power by appropriately integrating distributed power sources and conventional power distribution equipment. In this paper, we introduce a battery storage to the power distribution system including road heating, and extend the predictive switching control of the systems due to the authors' previous study to the case where battery storage is installed. As a main result, we propose a predictive switching control that utilizes photovoltaic (PV) power generation and surplus power stored in the battery storage effectively, and achieves the reduction of distribution loss, attenuation of voltage fluctuation, and efficient snow melting, simultaneously. We verify the effectiveness of the application of battery storage through numerical simulation using actual time series data of weather conditions and active power of the PV power generation and load.

Recently, chaotic phenomena in laser dynamics have attracted much attention to its applied aspects, and a synchronization phenomenon, leader-laggard relationship, in time-delay coupled lasers has been used in reinforcement learning. In the present paper, we discuss the possibility of capturing the essential stochasticity of the leader-laggard relationship; in nonlinear science, it is known that coarse-graining allows one to derive stochastic models from deterministic systems. We derive stochastic models with the aid of the Koopman operator approach, and we clarify that the low-pass filtered data is enough to recover the essential features of the original deterministic chaos, such as peak shifts in the distribution of being the leader and a power-law behavior in the distribution of switching-time intervals. We also confirm that the derived stochastic model works well in reinforcement learning tasks, i.e., multi-armed bandit problems, as with the original laser chaos system.