Search

Mean-field games (MFGs) are developed to model the decision-making processes of a large number of interacting agents in multi-agent systems. This paper studies mean-field games on graphs (G -MFGs). The equilibria of G -MFGs, namely, mean-field equilibria (MFE), are challenging to solve for their high-dimensional action space because each agent has to make decisions when they are at junction nodes or on edges. Furthermore, when the initial population state varies on graphs, we have to recompute MFE, which could be computationally challenging and memory-demanding. To improve the scalability and avoid repeatedly solving G -MFGs every time their initial state changes, this paper proposes physics-informed graph neural operators (PIGNO). The PIGNO utilizes a graph neural operator to generate population dynamics, given initial population distributions. To better train the neural operator, it leverages physics knowledge to propagate population state transitions on graphs. A learning algorithm is developed, and its performance is evaluated on autonomous driving games on road networks. Our results demonstrate that the PIGNO is scalable and generalizable when tested under unseen initial conditions. [ABSTRACT FROM AUTHOR]

This paper proposes a scalable learning framework to solve a system of coupled forward–backward partial differential equations (PDEs) arising from mean field games (MFGs). The MFG system incorporates a forward PDE to model the propagation of population dynamics and a backward PDE for a representative agent's optimal control. Existing work mainly focus on solving the mean field game equilibrium (MFE) of the MFG system when given fixed boundary conditions, including the initial population state and terminal cost. To obtain MFE efficiently, particularly when the initial population density and terminal cost vary, we utilize a physics-informed neural operator (PINO) to tackle the forward–backward PDEs. A learning algorithm is devised and its performance is evaluated on one application domain, which is the autonomous driving velocity control. Numerical experiments show that our method can obtain the MFE accurately when given different initial distributions of vehicles. The PINO exhibits both memory efficiency and generalization capabilities compared to physics-informed neural networks (PINNs). [ABSTRACT FROM AUTHOR]

Topological phases of matter are conventionally characterized by the bulk‐boundary correspondence in Hermitian systems. The topological invariant of the bulk in d dimensions corresponds to the number of (d − 1)‐dimensional boundary states. By extension, higher‐order topological insulators reveal a bulk‐edge‐corner correspondence, such that nth order topological phases feature (d − n)‐dimensional boundary states. The advent of non‐Hermitian topological systems sheds new light on the emergence of the non‐Hermitian skin effect (NHSE) with an extensive number of boundary modes under open boundary conditions. Still, the higher‐order NHSE remains largely unexplored, particularly in the experiment. An unsupervised approach—physics‐graph‐informed machine learning (PGIML)—to enhance the data mining ability of machine learning with limited domain knowledge is introduced. Through PGIML, the second‐order NHSE in a 2D non‐Hermitian topoelectrical circuit is experimentally demonstrated. The admittance spectra of the circuit exhibit an extensive number of corner skin modes and extreme sensitivity of the spectral flow to the boundary conditions. The violation of the conventional bulk‐boundary correspondence in the second‐order NHSE implies that modification of the topological band theory is inevitable in higher dimensional non‐Hermitian systems. [ABSTRACT FROM AUTHOR]

Composite materials have been widely used in the aerospace industry and are critical for safe operations. However, the delamination, caused by cyclic loads and corrosive service environment, poses a serious threat to the structural integrity of composite laminates. The acoustic emission technique has been adopted to assess the structural integrity by characterizing damage location, type, and size. This article proposes and compares data-driven prognostic methods to quantify the delamination area efficiently and accurately. To address the problem of insufficient inspection data, the prediction model adopts the path length across the delamination area as the target value. The delamination area can then be estimated with the predicted path length based on formulated geometric relationships. This solution will augment the model training datasets, and consequently, avoid the overfitting problem during the training process. Experimental results on composite coupons demonstrate that the proposed ensemble learning-based model outperforms other state-of-the-art methods in terms of prediction accuracy and efficiency. [ABSTRACT FROM AUTHOR]

• This work proposes a machine learning aided nanostructure reconstruction method. • The surrogate electromagnetic solver predicts the signatures fast and accurately. • A signature dimensionality reduction approach improves the computational efficiency. • The MLER achieves fast extraction compared to conventioanl library search method. • The scale of dataset for the implementation is smaller than conventional method. Optical scatterometry is the workhorse technique for in-line manufacturing process control in the semiconductor industry. However, as manufacturing processes develop, traditional methods for solving the inverse problem in optical scatterometry are struggling to continue improving productivity. To address this problem, machine learning can be a promising method, but it is a challenge to ensure robustness. In this paper, we propose a machine learning method to reconstruct the profile of nanostructures. The proposed method consists of three parts: compressing signature using a dimensionality reduction approach based on the principle component analysis, constructing a surrogate electromagnetic solver (SurEM) based on an artificial neural network mapping from parameters to signatures, and iteratively comparing the SurEM-predicted signatures with measured one to finally determine the results. Experiments have demonstrated that the proposed method can achieve fast and accurate measurement. This method is thus promising as an efficient in-line measurement method for nano- or micro-scale manufacturing. [ABSTRACT FROM AUTHOR]

The load-bearing capacity of damaged composite materials can be improved using the double adhesive repair technique. This study uses acoustic emission (AE) as the primary detection method to examine the damage evolution of modified double-bonded repair specimens and discover the most effective configuration for bi-adhesive repair. The introduction of machine learning and signal feature extraction in the analysis of huge acoustic emission parameters has significantly improved the accuracy and efficiency of damage classification. Following the preliminary division of the clustering results of the unsupervised algorithm, the corresponding relationship between various clustering types and damage types is confirmed using the energy characteristics of wavelet packet decomposition. Parameter and waveform analysis are extensively utilized in studying the AE damage of composite materials and have complementary advantages. Thus, a supervised classifier is established by combining characteristic parameters with time-frequency domain analysis of the signal waveform. Furthermore, machine learning is used to process the large signal set and analyze the debonding damage behavior of the patch. In the end, it is shown that the specimens with the modified adhesive in the middle of the repair area displayed superior repair results. Hence, when combined with machine learning, the signal classification of the damage evolution behavior of repaired specimens is realized with high accuracy. [ABSTRACT FROM AUTHOR]

Change detection determines and evaluates changes by comparing bi-temporal images, which is a challenging task in the remote-sensing field. To better exploit the high-level features, deep-learning-based change-detection methods have attracted researchers' attention. Most deep-learning-based methods only explore the spatial–spectral features simultaneously. However, we assume the key spatial-change areas should be more important, and attention should be paid to the specific bands which can best reflect the changes. To achieve this goal, we propose the spatial–spectral joint attention network (SJAN). Compared with traditional methods, SJAN introduces the spatial–spectral attention mechanism to better explore the key changed areas and the key separable bands. To be more specific, a novel spatial-attention module is designed to extract the spatially key regions first. Secondly, the spectral-attention module is developed to adaptively focus on the separable bands of land-cover materials. Finally, a novel objective function is proposed to help the model to measure the similarity of learned spatial–spectral features from both spectrum amplitude and angle perspectives. The proposed SJAN is validated on three benchmark datasets. Comprehensive experiments have been conducted to demonstrate the effectiveness of the proposed SJAN. [ABSTRACT FROM AUTHOR]

Pulsed eddy current (PEC) technique has been applied as a viable method to detect hidden discontinuities in metallic structures. Conventionally, selected time-domain features are employed to characterize the PEC data, such as peak value, lift-off point of intersection, rising point, crossing time, and differential time to peak. The research presented in this paper continues the effort in a previous study on detecting the radial cracks starting from the fastener hole in second layer of a two-layer mock-up aircraft structure. A large diameter excitation coil with ferrite core is used to induce a strong pulse, and the magnetic field generated by eddy current is detected by Hall sensors. Instead of analyzing the limited time-domain features, we propose using machine learning methods to interpret the raw data without feature extraction. Thus, the second-layer discontinuities can be characterized presumably with all the information contained in a waveform. An automated detection framework is proposed in this paper and the experimental results demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

The fundamental transmission channels of smart wearable devices and systems should support function reconfiguration and environment cognition or self‐adaption. Here, a environment‐adaptive meta‐channel (EAMC) is proposed that is adaptive to the near‐field environment. Benefiting from the capability to manipulate environment‐wave interactions, EAMC can cognize the near‐field environment impurities with subwavelength resolution, in which different impurities can be distinguished and located accurately with the aid of machine learning method. Thus, EAMC can adapt itself to varying environment so that its customized function is stabilized. For instance, EAMC is used as a communication modulator to adapt to different dielectric impurities and human body contacts dynamically. The quality of transmitted picture signals with impurities will deteriorate seriously but can be repaired efficiently as the self‐adaptive process of EAMC is used. Hence, the proposed EAMC could be a candidate for fundamental transmission channels for future smart wearable devices and systems. [ABSTRACT FROM AUTHOR]

The internal structure of wind turbines is intricate and precise, although the challenging working conditions often give rise to various operational faults. This study aims to address the limitations of traditional machine learning algorithms in wind turbine fault detection and the imbalance of positive and negative samples in the fault detection dataset. To achieve the real-time detection of wind turbine group faults and to capture wind turbine fault state information, an enhanced ASL-CatBoost algorithm is proposed. Additionally, a crawling animal search algorithm that incorporates the Tent chaotic mapping and t-distribution mutation strategy is introduced to assess the sensitivity of the ASL-CatBoost algorithm toward hyperparameters and the difficulty of manual hyperparameter setting. The effectiveness of the proposed hyperparameter optimization strategy, termed the TtRSA algorithm, is demonstrated through a comparison of traditional intelligent optimization algorithms using 11 benchmark test functions. When applied to the hyperparameter optimization of the ASL-CatBoost algorithm, the TtRSA-ASL-CatBoost algorithm exhibits notable enhancements in accuracy, recall, and other performance measures compared with the ASL-CatBoost algorithm and other ensemble learning algorithms. The experimental results affirm that the proposed algorithm model improvement strategy effectively enhances the wind turbine fault detection classification recognition rate. [ABSTRACT FROM AUTHOR]

Background. Nurses' high workload can result in depressive symptoms. However, the research has underexplored the internal and external variables, such as organisational support, career identity, and burnout, which may predict depressive symptoms among Chinese nurses via machine learning (ML). Aim. To predict nurses' depressive symptoms and identify the relevant factors by machine learning (ML) algorithms. Methods. A self-administered smartphone questionnaire was delivered to nurses to evaluate their depressive symptoms; 1,431 questionnaires and 28 internal and external features were collected. In the training set, the use of maximum relevance minimum redundancy ranked the features' importance. Five ML algorithms were used to establish models to identify nurses' depressive symptoms using different feature subsets, and the area under the curve (AUC) determined the optimal feature subset. Demographic characteristics were added to the optimal feature subset to establish the combined models. Each model's performance was evaluated using the test set. Results. The prevalence rate of depressive symptoms among Chinese nurses was 31.86%. The optimal feature subset comprised of sleep disturbance, chronic fatigue, physical fatigue, exhaustion, and perceived organisation support. The five models based on the optimal feature subset had good prediction performance on the test set (AUC: 0.871–0.895 and accuracy: 0.798–0.815). After adding the significant demographic characteristics, the performance of the five combined models slightly improved; the AUC and accuracy increased to 0.904 and 0.826 on the test set, respectively. The logistic regression analysis results showed the best and most stable performance while the univariate analysis results showed that external and internal personal features (AUC: 0.739–0.841) were more effective than demographic characteristics (AUC: 0.572–0.588) for predicting nurses' depressive symptoms. Conclusions. ML could effectively predict nurses' depressive symptoms. Interventions to manage physical fatigue, sleep disorders, burnout, and organisational support may prevent depressive symptoms. [ABSTRACT FROM AUTHOR]

This study aimed to investigate the association between long-term exposure to fine particulate matter (PM 2.5) and its constituents (black carbon (BC), ammonium (NH 4 +), nitrate (NO 3 −), organic matter (OM), inorganic sulfate (SO 4 2−)) and incident female breast cancer in Beijing, China. Data from a prospective cohort comprising 85,504 women enrolled in the National Urban Cancer Screening Program in Beijing (2013–2019) and the Tracking Air Pollution in China dataset are used. Monthly exposures were aggregated to calculate 5-year average concentrations to indicate long-term exposure. Cox models and mixture exposure models (weighted quantile sum, quantile-based g-computation, and explanatory machine learning model) were employed to analyze the associations. Findings indicated increased levels of PM 2.5 and its constituents were associated with higher breast cancer risk, with hazard ratios per 1-μg/m3 increase of 1.02 (95% confidence interval (CI): 1.01, 1.03), 1.39 (95% CI: 1.16, 1.65), 1.28 (95% CI: 1.12, 1.46), 1.15 (95% CI: 1.05, 1.24), 1.05 (95% CI: 1.02, 1.08), and 1.15 (95% CI: 1.07, 1.23) for PM 2.5 , BC, NH 4 +, NO 3 −, OM, and SO 4 2−, respectively. Exposure-response curves demonstrated a monotonic risk increase without an evident threshold. Mixture exposure models highlighted BC and SO 4 2− as key factors, underscoring the importance of reducing emissions of these pollutants. [Display omitted] • Long-term exposure to PM 2.5 and its constituents was linked to higher female breast cancer risk in Beijing. • Key pollutants black carbon (BC) and sulfate (SO 4 2−) were identified as significant contributors to breast cancer risk. • Effective control measures are crucial in mitigating pollutants to reduce the risk of female breast cancer. [ABSTRACT FROM AUTHOR]

• A channeled spectroscopic ellipsometry (CSE) enabled by the physics-informed untrained neural networks (PITUNN) is proposed to deal with systematic errors and random noises in measurement. • The PITUNN needs no prior dataset or training, demonstrating robust generalization capabilities in principle. • The PITUNN employs simple networks in tandem, showing rapid performance with high efficiency. • The simulated and experimental results demonstrate that CSE enabled by the PITUNN exhibits high measurement accuracy and robustness. Ellipsometry is a powerful metrology technique for characterizing the optical properties of various materials. Channeled spectroscopic ellipsometry (CSE) has shown great promise among the different types of ellipsometry due to its simple setup and rapid performance. Furthermore, CSE modulates the polarization parameters of thin films into a spectrum, thus transforming the measurement process into a demodulation problem. However, conventional CSE faces challenges in measurement accuracy and computational efficiency, with strict hardware and calibration requirements. Inspired by physics-informed machine learning, we propose CSE enabled by the physics-informed tandem untrained neural networks (PITUNN), which does not require training, exhibits high computational efficiency and partially alleviates the strict requirements for hardware and calibration accuracy. We also demonstrate the effectiveness of CSE enabled by the PITUNN and its ability to handle system errors and random noise through simulations and experiments on thin films of different thickness and materials. [ABSTRACT FROM AUTHOR]

• Disability level (assessed by six-minute walk test) in people with multiple sclerosis was estimated using the data collected from wearable devices in free-living conditions. • Fitbit minute-level step counts were analysed from which features were extracted. • Elastic net, gradient boosted trees and random forest were employed to build machine learning models. • Promising estimation and classification performances were obtained. • Features capturing the upper end of the step count distribution showed greater importance. Multiple sclerosis (MS) is a progressive inflammatory and neurodegenerative disease of the central nervous system affecting over 2.5 million people globally. In-clinic six-minute walk test (6MWT) is a widely used objective measure to evaluate the progression of MS. Yet, it has limitations such as the need for a clinical visit and a proper walkway. The widespread use of wearable devices capable of depicting patients' activity profiles has the potential to assess the level of MS-induced disability in free-living conditions. In this work, we extracted 96 features in different temporal granularities (from minute-level to day-level) from wearable data and explored their utility in estimating 6MWT scores in a European (Italy, Spain, and Denmark) MS cohort of 337 participants over an average of 10 months' duration. We combined these features with participants' demographics using three regression models including elastic net, gradient boosted trees and random forest. In addition, we quantified the individual feature's contribution using feature importance in these regression models, linear mixed-effects models, generalized estimating equations, and correlation-based feature selection (CFS). The results showed promising estimation performance with R2 of 0.30, which was derived using random forest after CFS. This model was able to distinguish the participants with low disability from those with high disability. Furthermore, we observed that the minute-level (≤ 8 minutes) step count, particularly those capturing the upper end of the step count distribution, had a stronger association with 6MWT. The use of a walking aid was indicative of ambulatory function measured through 6MWT. This study demonstrates the utility of wearables devices in assessing ambulatory impairments in people with MS in free-living conditions and provides a basis for future investigation into the clinical relevance. [ABSTRACT FROM AUTHOR]