Yield curve extrapolation to unobservable tenors is a key technique for the market-consistent valuation of actuarial liabilities required by Solvency II and forthcoming similar regulations. Since the regulatory method, the Smith–Wilson method, is inconsistent with observable yield curve dynamics, parsimonious parametric models, the Nelson–Siegel model and its extensions, are often used for yield curve extrapolation in risk management. However, it is difficult for the parsimonious parametric models to extrapolate yield curves without excessive volatility because of their limited ability to represent observed yield curves with a limited number of parameters. To extend the representational capabilities, we propose a novel yield curve extrapolation method using machine learning. Using the long short-term memory architecture, we achieve purely data-driven yield curve extrapolation with better generalization performance, stability, and consistency with observed yield curve dynamics than the previous parsimonious parametric models on US and Japanese yield curve data. In addition, our method has model interpretability using the backpropagation algorithm. The findings of this study prove that neural networks, which have recently received considerable attention in mortality forecasting, are useful for yield curve extrapolation, where they have not been used before.

In order to improve the global convergence performance of the super-twisting sliding mode control (STSMC) for the uncertain hybrid mechanism, especially in the high-speed scenario, and enhance the robustness of hybrid mechanism system to the uncertainties with a wide range of changes, an intelligent fixed-time super-twisting sliding mode control (IFTSTSMC) is proposed. Firstly, a fixed-time super-twisting sliding mode control (FTSTSMC) algorithm is designed by adding the exponential power terms with the fixed-time performance parameters in sliding variables and the transcendental function of the super-twisting algorithm in order to enhance the global convergence performance of the STSMC. Secondly, the existence condition of FTSTSMC for the uncertain hybrid mechanism is analyzed. The IFTSTSMC is designed by introducing RBF neural network to break through the limited range of uncertainties in FTSTSMC and enhance the robustness of hybrid mechanism system to the uncertainties with a wide range of changes. Then, the Lyapunov stability of the proposed method and the global fixed-time convergence of the system are proved theoretically. Finally, the effectiveness and superiority of the proposed control method are verified by the simulation and the automobile electro-coating conveying prototype experiment comparing with two classical finite-time sliding mode control methods.

Liquid crystal microwave phased arrays (LC-MPAs) are regarded as an ideal approach to realize compact antennas owing to their advantages in cost, size, weight, and power consumption. However, the shortcoming in low radiation deflection efficiency has been one of LC-MPAs’ main application limitations. To optimize the steering performance of LC-MPAs, it is essential to model the channel imperfections and compensate for the phase errors. In this paper, a phase error estimation model is built by training a neural network to establish a nonlinear relationship between the near-field phase error and the far-field pattern, hence realizing fast calibration for LC-MPAs within several measured patterns. Simulations and experiments on a 64-channel, two-dimensional planar antenna were conducted to validate this method. The results show that this method offers precise phase error estimations of 3.58° on average, realizes a fast calibration process with several field-measured radiation patterns, and improves the performances of the LC-MPA by approximately 4%–10% in deflection efficiency at different steering angles.

Motion and constraint identification are the fundamental issue throughout the development of parallel mechanisms. Aiming at meaningful result with heuristic and visualizable process, this paper proposes a machine learning-based method for motions and constraints modeling and further develops the automatic software for mobility analysis. As a preliminary, topology of parallel mechanism is characterized by recognizable symbols and mapped to the motion of component limb through programming algorithm. A predictive model for motion and constraint with their nature meanings is constructed based on neural network. An increase in accuracy is obtained by the novel loss function, which combines the errors of network and physical equation. Based on the predictive model, an automatic framework for mobility analysis of parallel mechanisms is constructed. A software is developed with WebGL interface, providing the result of mobility analysis as well as the visualizing process particularly. Finally, five typical parallel mechanisms are taken as examples to verify the approach and its software. The method facilitates to attain motion/constraint and mobility of parallel mechanisms with both numerical and geometric features.

In this paper, we construct interpretable zero-inflated neural network models for modeling hospital admission counts related to respiratory diseases among a health-insured population and their dependants in the United States. In particular, we exemplify our approach by considering the zero-inflated Poisson neural network (ZIPNN), and we follow the combined actuarial neural network (CANN) approach for developing zero-inflated combined actuarial neural network (ZIPCANN) models for modeling admission rates, which can accommodate the excess zero nature of admission counts data. Furthermore, we adopt the LocalGLMnet approach (Richman & Wüthrich (2023). Scandinavian Actuarial Journal, 2023(1), 71–95.) for interpreting the ZIPNN model results. This facilitates the analysis of the impact of a number of socio-demographic factors on the admission rates related to respiratory disease while benefiting from an improved predictive performance. The real-life utility of the methodologies developed as part of this work lies in the fact that they facilitate accurate rate setting, in addition to offering the potential to inform health interventions.

Plasma-enhanced atomic layer deposition (PEALD) is gaining interest in thin films for laser applications, and post-annealing treatments are often used to improve thin film properties. However, research to improve thin film properties is often based on an expensive and time-consuming trial-and-error process. In this study, PEALD-HfO2 thin film samples were deposited and treated under different annealing atmospheres and temperatures. The samples were characterized in terms of their refractive indices, layer thicknesses and O/Hf ratios. The collected data were split into training and validation sets and fed to multiple back-propagation neural networks with different hidden layers to determine the best way to construct the process–performance relationship. The results showed that the three-hidden-layer back-propagation neural network (THL-BPNN) achieved stable and accurate fitting. For the refractive index, layer thickness and O/Hf ratio, the THL-BPNN model achieved accuracy values of 0.99, 0.94 and 0.94, respectively, on the training set and 0.99, 0.91 and 0.90, respectively, on the validation set. The THL-BPNN model was further used to predict the laser-induced damage threshold of PEALD-HfO2 thin films and the properties of the PEALD-SiO2 thin films, both showing high accuracy. This study not only provides quantitative guidance for the improvement of thin film properties but also proposes a general model that can be applied to predict the properties of different types of laser thin films, saving experimental costs for process optimization.

Bees play a significant role in the health of terrestrial ecosystems. The decline of bee populations due to colony collapse disorder around the world constitutes a severe ecological danger. Maintaining high yield of honey and understanding of bee behaviour necessitate constant attention to the hives. Research initiatives have been taken to establish monitoring programs to study the behaviour of bees in accessing their habitat. Monitoring the sanitation and development of bee brood allows for preventative measures to be taken against mite infections and an overall improvement in the brood's health. This study proposed a precision beekeeping method that aims to reduce bee colony mortality and improve conventional apiculture through the use of technological tools to gather, analyse, and understand bee colony characteristics. This research presents the application of advanced digital image processing with computer vision techniques for the visual identification and analysis of bee brood at various developing stages. The beehive images are first preprocessed to enhance the important features of object. Further, object is segmented and classified using computer vision techniques. The research is carried out with the images containing variety of immature brood stages. The suggested method and existing methods are tested and compared to evaluate efficiency of proposed methodology.

Predicting reservoir storage capacities is an important planning activity for effective conservation and water release practices. Weather events such as drought and precipitation impact water storage capacities in reservoirs. Predictive insights on reservoir storage levels are beneficial for water planners and stakeholders in effective water resource management. A deep learning (DL) neural network (NN) based reservoir storage prediction approach is proposed that learns from climate, hydrological, and storage information within the reservoir’s associated watershed. These DL models are trained and evaluated for 17 reservoirs in Texas, USA. Using the trained models, reservoir storage predictions were validated with a test data set spanning 2 years. The reported results show promise for longer-term water planning decisions.

A synthetic flow angle sensor, able to estimate angle-of-attack and angle-of-sideslip, can exploit different methods to solve a set of equations modelling data fusion from other onboard systems. In operative scenarios, measurements used for data fusion are characterised by several uncertainties that would significantly affect the synthetic sensor performance. The off-line use of neural networks is not a novelty to model deterministic synthetic flow angle sensors and to mitigate issues arising from real flight applications. A common practice is to train the neural network with corrupted data that are representative of uncertainties of the current application. However, this approach requires accurate tuning on the target aircraft and extensive flight test campaigns, therefore, making the neural network tightly dependent on the specific aircraft. In order to overcome latter issues, this work proposes the use of neural networks to solve a model-free scheme, derived from classical flight mechanics, that is independent from the target aircraft, flight regime and avionics. It is crucial to make use of a training dataset that is not related to any specific aircraft or avionics to preserve the generality of the scheme. Under these circumstances, global and local neural networks are herein compared with an iterative method to assess the neural capabilities to generalise the proposed model-free solver. The final objective of the present work, in fact, is to select the neural technique that can enable a flow angle synthetic sensor to be used on board any flying body at any flight regime without any further training sessions.

A domain-theoretic framework is presented for validated robustness analysis of neural networks. First, global robustness of a general class of networks is analyzed. Then, using the fact that Edalat’s domain-theoretic L-derivative coincides with Clarke’s generalized gradient, the framework is extended for attack-agnostic local robustness analysis. The proposed framework is ideal for designing algorithms which are correct by construction. This claim is exemplified by developing a validated algorithm for estimation of Lipschitz constant of feedforward regressors. The completeness of the algorithm is proved over differentiable networks and also over general position ${\mathrm{ReLU}}$ networks. Computability results are obtained within the framework of effectively given domains. Using the proposed domain model, differentiable and non-differentiable networks can be analyzed uniformly. The validated algorithm is implemented using arbitrary-precision interval arithmetic, and the results of some experiments are presented. The software implementation is truly validated, as it handles floating-point errors as well.

Collaborative planning for multiple hypersonic vehicles can effectively improve operational effectiveness. Time coordination is one of the main forms of cooperation among multi-hypersonic glide vehicles, and time cooperation trajectory optimisation is a key technology that can significantly increase the success rate of flight missions. However, it is difficult to obtain satisfactory time as a constraint condition during trajectory optimisation. To solve this problem, a multilayer Perceptrona is trained and adopted in a time-decision module, whose input is a four-dimensional vector selected according to the trajectory characteristics. Additionally, the MLP will be capable of determining the optimal initial heading angle of each aircraft to reduce unnecessary manoeuvering performance consumption in the flight mission. Subsequently, to improve the cooperative flight performance of hypersonic glide vehicles, the speed-dependent angle-of-attack and bank command were designed and optimised using the Artificial Bee Colony algorithm. The final simulation results show that the novel strategy proposed in this study can satisfy terminal space constraints and collaborative time constraints simultaneously. Meanwhile, each aircraft saves an average of 13.08% flight range, and the terminal speed is increased by 315.6m/s compared to the optimisation results of general purpose optimal control software (GPOPS) tools.

A multiple-vehicles time-coordination guidance technique based on deep learning is suggested to address the cooperative guiding problem of hypersonic gliding vehicle entry phase. A dual-parameter bank angle profile is used in longitudinal guiding to meet the requirements of time coordination. A vehicle trajectory database is constructed along with a deep neural network (DNN) structure devised to fulfill the error criteria, and a trained network is used to replace the conventional prediction approach. Moreover, an extended Kalman filter is constructed to detect changes in aerodynamic parameters in real time, and the aerodynamic parameters are fed into a DNN. The lateral guiding employs a logic for reversing the sign of bank angle, which is based on the segmented heading angle error corridor. The final simulation results demonstrate that the built DNN is capable of addressing the cooperative guiding requirements. The algorithm is highly accurate in terms of guiding, has a fast response time, and does not need inter-munition communication, and it is capable of solving guidance orders that satisfy flight requirements even when aerodynamic parameter disruptions occur.

Plasma vertical displacement control is essential for the stable operation of tokamak devices. The traditional plasma vertical displacement calculation method is not suitable for balancing speed and accuracy simultaneously, which is necessary for real-time feedback control. In this study, neural networks are used to rapidly detect vertical displacement recognition. Based on a fully connected neural network, the vertical displacement calculation model is trained and tested using magnetic data of approximately 2000 shots. To compare the effects of different inputs on vertical displacement calculation, different magnetic measurement diagnostic signals are used to train and test the model. Compared with a full magnetic measurement dataset, 39 magnetic measurement signals (38 magnetic probes and plasma current) show better accuracy with mean square error <0.0005. The model is tested using historical experimental data, and it demonstrates accurate vertical displacement calculation even in the case of a vertical displacement event. In general, neural network algorithm has great application potential in vertical displacement calculation.

Understanding various historical entity information (e.g., persons, locations, and time) plays a very important role in reasoning about the developments of historical events. With the increasing concern about the fields of digital humanities and natural language processing, named entity recognition (NER) provides a feasible solution for automatically extracting these entities from historical texts, especially in Chinese historical research. However, previous approaches are domain-specific, ineffective with relatively low accuracy, and non-interpretable, which hinders the development of NER in Chinese history. In this paper, we propose a new hybrid deep learning model called “subword-based ensemble network” (SEN), by incorporating subword information and a novel attention fusion mechanism. The experiments on a massive self-built Chinese historical corpus CMAG show that SEN has achieved the best with 93.87% for F1-micro and 89.70% for F1-macro, compared with other advanced models. Further investigation reveals that SEN has a strong generalization ability of NER on Chinese historical texts, which is not only relatively insensitive to the categories with fewer annotation labels (e.g., OFI) but can also accurately capture diverse local and global semantic relations. Our research demonstrates the effectiveness of the integration of subword information and attention fusion, which provides an inspiring solution for the practical use of entity extraction in the Chinese historical domain.

Chronic food insecurity remains a challenge globally, exacerbated by climate change-driven shocks such as droughts and floods. Forecasting food insecurity levels and targeting vulnerable households is apriority for humanitarian programming to ensure timely delivery of assistance. In this study, we propose to harness a machine learning approach trained on high-frequency household survey data to infer the predictors of food insecurity and forecast household level outcomes in near real-time. Our empirical analyses leverage the Measurement Indicators for Resilience Analysis (MIRA) data collection protocol implemented by Catholic Relief Services (CRS) in southern Malawi, a series of sentinel sites collecting household data monthly. When focusing on predictors of community-level vulnerability, we show that a random forest model outperforms other algorithms and that location and self-reported welfare are the best predictors of food insecurity. We also show performance results across several neural networks and classical models for various data modeling scenarios to forecast food security. We pose that problem as binary classification via dichotomization of the food security score based on two different thresholds, which results in two different positive class to negative class ratios. Our best performing model has an F1 of 81% and an accuracy of 83% in predicting food security outcomes when the outcome is dichotomized based on threshold 16 and predictor features consist of historical food security score along with 20 variables selected by artificial intelligence explainability frameworks. These results showcase the value of combining high-frequency sentinel site data with machine learning algorithms to predict future food insecurity outcomes.