Error Model Research Articles

Seq2Seq model with attention for predicting nonlinear propagation of ultrafast pulses in optical fibers

The propagation of ultrashort pulses in optical fibers exhibits highly complex nonlinear dynamics, which plays a central role in the development of light sources and photonic technologies. The mainstream of the existing methods for modeling and predicting complex nonlinear propagation dynamics of ultrashort pulses in optical fibers is based on recurrent neural networks (RNNs), which use a Multi-In-Single-Out (MISO) architecture to predict the optical pulse evolution recursively. This autoregressive model is severely limited by the error accumulation problem and also requires significant computational resources. Affected by the error accumulation problem, this method often leads to severe performance degradation in long sequence prediction tasks, thus limiting the practical application of the prediction model. In this work, we propose a new non-autoregressive model using a Single-In-Multi-Out (SIMO) architecture to simulate the highly nonlinear dynamics of ultrashort pulse propagation in optical fibers. Our model is validated on the public dataset. The results show that our model can significantly reduce the prediction error in modeling and predicting the complex nonlinear propagation of ultrashort pulses in optical fibers. In addition, the required computational resources and time spent are significantly reduced. As a whole, our proposed method comprehensively outperforms the mainstream methods in terms of efficiency, accuracy and practicality. We believe our work could bring new insights into the modeling and analysis of complex ultrafast nonlinear dynamics.

Hierarchical Bayesian quantification of aerodynamic effects on an offshore wind turbine under varying environmental and operational conditions

This paper proposes a hierarchical Bayesian model updating approach to quantify variability of aerodynamic stiffness and damping of an offshore wind turbine (OWT) under different environmental and operational conditions (EOCs) using in-situ vibration data and SCADA (supervisory control and data acquisition) over two months of continuous monitoring. The considered OWT is a Haliade 150, 6 MW GE turbine on a jacket substructure located at the Block Island Wind Farm in Rhode Island, USA. The OWT has been instrumented with a continuous monitoring system including an array of accelerometers and strain gauges. The modal parameters of the OWT are extracted using an automated system identification approach. These parameters exhibit significant variations under varying EOCs. This variation is more significant in natural frequencies and damping ratios of the first fore-aft bending mode due to the aerodynamic effects. In this paper, a modeling approach is proposed by introducing a spring and a damper at the nacelle level in the fore-aft direction to account for the observed aerodynamic effects. A hierarchical Bayesian model updating is formulated and implemented to update parameters representing the effects of aerodynamic stiffness and damping, as well as their statistical properties such as mean and covariance matrix which are updated as hyperparameters. To account for the correlation between aerodynamic effects and EOCs, the updating parameters are assumed to be functions of EOCs such as rpm and wind speed. Two levels of hierarchical Bayesian model updating are performed and compared. In level 1, only modal parameters are used in model updating, while in level 2, the assumed correlation between EOCs and modal parameters are accounted for to reduce the uncertainty of aerodynamic effects in the model. In addition to hyperparameters, the proposed hierarchical Bayesian approach provides statistical properties of error function by collecting the residual uncertainties that have not been accounted for, e.g., modeling errors, measurement noise and random disturbances. The predicted modal parameters using two levels of hierarchical Bayesian approach are compared with their identified counterparts, and the results indicate that level 2 approach significantly reduces estimation uncertainty in aerodynamic effects and produces model predictions consistent with observed values of natural frequencies and damping ratios.

Numerical investigation on the error of volume-averaged reaction rate model for the low-velocity filtration combustion with hybrid-scale method

An innovative “hybrid-scale” numerical simulation method was proposed to address the low-velocity filtration combustion within double-layer pellet-packed beds. This approach integrates the volume averaging method in the preheating and post-flame zones while direct pore-level simulation in the reaction zone. Based on the hybrid-scale simulation results, a comprehensive analysis was conducted on the theoretical rigor and errors arising from the application of the volume averaging method. The results suggest that in the strictest sense, the volume-averaging method is not suitable for the reaction zone, as the flame thickness is typically smaller than the characteristic length of a representative element volume (REV). However, when the characteristic length is comparable to the pellet diameter, the spatial deviation values of temperature and component concentration in the majority (>80 %) of the reaction zone are smaller than their intrinsic average values. The error introduced by the volume-averaged reaction rate model primarily originates from the spatial heterogeneity of the component concentration field within the REV. The existing volume-averaged reaction rate model tends to underestimate the actual reaction rate by up to 20 %. The recommended mesh size when using volume averaging method should be comparable to the equivalent diameter of the pellet to improve the calculation accuracy

Sequential multiscale simulation of heat transfer and experimental verification of porous phenolic resin composites under Knudsen effect

The randomness and multi-level structures inherent to porous composites with open-pore make it difficult to establish equivalent geometric models at different scales for multi-scale simulations. This paper presents a combined experimental and simulation approach to the preparation, structural analysis and multiscale simulation of quartz fibre-reinforced phenolic composites. The elementalized open-pore porous models with the Knudsen effect, the random fibre yarn models and the random fibre felt models have been established and assembled into a composite structural model after homogenization. The thermal conductivity parameters of the porous model are calculated and transferred to the fibre yarn and fibre felt models for simulation. Thereafter, the thermal conductivity parameters of the three models are transferred to the composite structure model and simulated to obtain its equivalent thermal conductivity. The experimental and simulation results demonstrate that the introduction of the Knudsen effect can reduce the simulation error of the composite structure model by an order of magnitude. In combination with the random contact characteristics of the yarns, the sequential multiscale finite element heat transfer simulation with an error of 0.5% can be achieved.

Methods for detecting, building, and improving tryptophan mannosylation in glycoprotein structures.

Tryptophan mannosylation, the covalent addition of an α-ᴅ-mannose sugar to a tryptophan side chain, is a post-translational modification (PTM) that can affect protein stability, folding, and interactions. Compared to other forms of protein glycosylation, it is relatively uncommon but is affected by conformational anomalies and modeling errors similar to those seen in N- and O-glycans in the Protein Data Bank (PDB). In this work, we report methods for detecting, building, and improving mannose structures linked to tryptophans. These methods have been used to mine X-ray crystallographic and cryo-electron microscopy maps in the PDB looking for unmodeled mannosylation, resulting in a number of cases where the modification can be placed in the map with high confidence. Additionally, we address most conformational issues affecting this modification. Finally, the development of a structural template to recognize thrombospondin repeats (TSR) domains where tryptophan mannosylation occurs will allow for the mannosylation of candidate-predicted models, for example, those predicted with AlphaFold.

A Bayesian Approach to Correcting the Attenuation Bias of Regression Using Polygenic Risk Score.

Polygenic risk score (PRS) has become increasingly popular for predicting the value of complex traits. In many settings, PRS is used as a covariate in regression analysis to study the association between different phenotypes. However, measurement error in PRS causes attenuation bias in the estimation of regression coefficients. In this paper, we employ a Bayesian approach to accounting for the measurement error of PRS and correcting the attenuation bias in linear and logistic regression. Through simulation, we show that our approach is able to obtain approximately unbiased estimation of coefficients and credible intervals with correct coverage probability. We also empirically compare our Bayesian measurement error model to the conventional regression model by analyzing real traits in the UK Biobank. The results demonstrate the effectiveness of our approach as it significantly reduces the error in coefficient estimates.

A novel virtual reality fear conditioning paradigm to investigate the influence of expectancy violation on fear extinction

Exposure therapy is an efficient treatment for pathological anxiety, yet its underlying mechanisms are not fully understood. Prediction error models suggest that optimizing the violation of threat-related expectancies improves treatment outcomes, however, causal evidence is still sparse. The aim of the current study was therefore to provide causal evidence for the influence of the extent of expectancy violations on extinction retention using a novel virtual reality fear conditioning paradigm. In total, 100 participants completed a two-day fear conditioning paradigm in which the approach behavior of an animated stimulus was differentially reinforced with an electrical stimulus (i.e., closer distances were associated with a higher probability for receiving the aversive stimulation). To experimentally manipulate the extent of expectancy violations during fear extinction, participants were presented only with distances to the conditioned stimulus that either weakly (i.e., far distances) or strongly predicted the aversive outcome (i.e., close distances), resulting in low vs. high expectancy violations. We found successful fear acquisition and extinction, as well as an influence of the extent of expectancy violations on US-expectancy and threat ratings after extinction on day 1. On the second day, at a spontaneous recovery and reinstatement test, however, there was only weak evidence for improved extinction retention in the high expectancy violation condition. Optimizing expectancy violations might be a necessary but not sufficient condition for improved extinction retention. Future research needs to address the conditions under which expectancy violations lead to robust expectancy adjustments and how these conditions can be met in exposure therapy for anxiety disorders.

Smart home-assisted anomaly detection system for older adults: a deep learning approach with a comprehensive set of daily activities.

Smart homes have the potential to enable remote monitoring of the health and well-being of older adults, leading to improved health outcomes and increased independence. However, current approaches only consider a limited set of daily activities and do not combine data from individuals. In this work, we propose the use of deep learning techniques to model behavior at the population level and detect significant deviations (i.e., anomalies) while taking into account the whole set of daily activities (41). We detect and visualize daily routine patterns, train a set of recurrent neural networks for behavior modelling with next-day prediction, and model errors with a normal distribution to identify significant deviations while considering the temporal component. Clustering of daily routines achieves a silhouette score of 0.18 and the best model obtains a mean squared error in next day routine prediction of 4.38%. The mean number of deviated activities for the anomalies in the train and test set are 3.6 and 3.0, respectively, with more than 60% of anomalies involving three or more deviated activities in the test set. The methodology is scalable and can incorporate additional activities into the analysis.

The evaluation of asymmetrically aging effect on silicon rubber insulators flashover voltage

Purpose This study aims to examine the effect of asymmetric aging (120 and 270 degrees) and symmetrical aging on the performance of silicone rubber insulators. The effects of different aged surfaces, pollution levels and humidity on flashover voltage (FV) have been experimentally investigated. Based on the laboratory results, a mathematical model has been presented to estimate the FV of insulators. Design/methodology/approach According to insulators constituent materials, polymer insulators are affected by environmental conditions, specifically by ultraviolet (UV) radiation. To investigate the effect of aging and environmental conditions on the insulator, two types of silicone rubber insulators were aged under UV radiation. Environmental conditions have been created by changing humidity in five levels and pollution in three levels. All tests have been performed on two virgins, two full aged and four asymmetric-aged insulators. Also, according to the effect of each parameter on the FV, a mathematical model has been presented for it. Findings UV aging causes the destruction of the silicon rubber insulator. The occurrence of flashover is one of the consequences of insulation failure. The enlargement of the aged surface has a nonlinear relationship with the reduction of the FV. By considering the effect of the aged surface, the accuracy of FV modeling can be increased. The results obtained from the mathematical model have shown that the estimation of the FV in different environmental conditions has an error of less than 5%. However, the modeling error is more than 15% if the effect of the aged surface is not considered. Originality/value In this work, it is intended to investigate the effect of asymmetric aging on the FV of the insulator. The obtained results have shown that the asymmetric aging surface is an important factor in changing the harmonic components of the leakage current, condition monitoring indicators and FV.

P-1323. Forecasting the incidence of acute respiratory infections in Uzbekistan, 2022-2023

Abstract Background Acute respiratory infections (ARIs) are a significant cause of morbidity and mortality worldwide, and this is particularly evident in regions such as Uzbekistan within Central Asia. Forecasting ARIs allows public health authorities to anticipate and prepare for surges in cases. This includes ensuring that healthcare facilities have adequate resources, such as hospital beds, ventilators, medicines, and personal protective equipment (PPE). Therefore, we aimed to predict the incidence of ARIs in Uzbekistan and respond to public health threats in a timely and appropriate manner.Figure 1.Short-term prediction of incidence acute respiratory infections in Uzbekistan, 2022-2023 Methods We conducted a retrospective study using weekly data on the incidence of ARIs reported in Uzbekistan between January 2022 and November 2023. Two models were selected to predict the number of ARIs cases. The first model, based on prediction using the EpiNow2 package, was used for short-term prediction for 2 weeks, and the second model, from the nnetar package, based on neural networks for prediction of univariate time series, was used for prediction for the next 12 weeks. Data analysis was performed in R version 4.3.2.Figure 2.Long term prediction of incidence acute respiratory infections in Uzbekistan, 2022-2023 Results The short-term forecast for the following 2 weeks showed a possible increase in case numbers (rate of growth = 0.011 (95% CI 0.004 - 0.028)). The model predicts 2,310 (767 - 2,939) new cases by infection date. At the end of the forecast period, the model predicted 5,711 (2,486 - 39,623) new cases of ARIs per week . The long-term model predicted 217,469 (176,261 - 247,929) new infections over the next 12 weeks. At the same time, the peak incidence was predicted to occur in December 2023. In that month, the model predicted 19,817 (17,340 - 22,303) new infections per week. When evaluating the accuracy of the model, we found that the predicted values were lower than the actual number of cases (mean absolute percentage error (MAPE) of the first model = 0.68, MAPE of the second model = 0.39). Conclusion The selected models of forecasting make it possible to determine the possible number of cases, although they underestimate values of these cases. The country needs to collect data on risk factors to better predict the incidence of ARIs. However, short- and long-term forecasting models can help to adjust public health response. Disclosures All Authors: No reported disclosures

P-1259. Transferring knowledge from gentamicin pharmacokinetics in neonates to other aminoglycosides

Abstract Background Gentamicin (GM) is frequently prescribed in the neonatal population, and its pharmacokinetics (PK) in this population are therefore well-studied. However, recent findings of diminished GM efficacy against Pseudomonas spp. have prompted increased use of other aminoglycosides with less well-defined PK, which poses a challenge for model-informed precision dosing (MIPD). Although ultimately the goal for dose individualization will be drug-specific models for neonatal patients, as an interim measure, this study investigates the performance of GM models in neonates receiving amikacin (AK) and tobramycin (TO). Patient Characteristics. Numbers are represented as counts or median (range) as appropriate. GA: gestational age; PMA: post-menstrual age; WT: weight; CR: creatinine. Methods Neonatal patient data entered into the InsightRX Nova model-informed precision dosing (MIPD) platform during routine clinical care were de-identified and analyzed retrospectively. Patients were included if their postnatal age was < 90 days or their postmenstrual age was < 44 weeks, received at least one dose of AK or TO and provided at least one therapeutic drug monitoring sample (TDM) within 48 hours of an administered dose. For each patient and for each model, samples were grouped by dosing interval and iteratively used to predict future levels using Bayesian forecasting, mimicking clinical implementation of MIPD. Prediction error was assessed using root mean square error (RMSE), mean percent error (MPE), and Accuracy, defined as percent of samples within 20% of predicted values or correctly identifying a trough under 1 mg/L. RMSE for aminoglycoside models developed on amikacin, gentamicin or tobramycin data sets and evaluated on clinical amikacin and tobramycin data sets. The best native performing model for each clinical data set serves as the reference model & is listed first, with subsequent models ordered by best to worst performance. Asterisks denotes statistical significance relative to the reference model. Results Overall, 14 AK courses with 31 AK TDMs, and 48 TO courses with 63 TO TDMs were included in the analysis (Table 1). RMSE, MPE, and Accuracy are displayed in Figures 1-3. Data on TO peaks were excluded due to low sample size. Of the AK courses, most GM models outperformed available AK models for both peak and trough levels. Of the TO courses, GM models exhibited significantly improved RMSE and MPE values, as well as significantly higher rates of accuracy when predicting trough levels compared to those of the available TO model. Mean percent error for aminoglycoside models developed on amikacin, gentamicin or tobramycin data sets and evaluated on clinical amikacin and tobramycin data sets. The best native performing model for each clinical data set serves as the reference model & is listed first, with subsequent models ordered by best to worst performance. Asterisks denotes statistical significance relative to the reference model. Conclusion PK models developed in neonates receiving GM match or outperform drug-specific models in predicting AK and TO exposures in neonates. Further research is needed to better understand model performance for TO peak levels in neonates and clinical implications of cross-module model use. Accuracy for aminoglycoside models developed on amikacin, gentamicin or tobramycin data sets and evaluated on clinical amikacin and tobramycin data sets. The best native performing model for each clinical data set serves as the reference model & is listed first, with subsequent models ordered by best to worst performance. Asterisks denotes statistical significance relative to the reference model. Disclosures Tiffany Lee, PharmD, InsightRX: Employee|InsightRX: Stocks/Bonds (Private Company) Jasmine Hughes, PhD, InsightRX: Employee|InsightRX: Stocks/Bonds (Private Company) Jon Faldasz, PharmD, BCPS, InsightRX: Employee|InsightRX: Stocks/Bonds (Private Company)

Analysis of past and future temperature variability and change in Southern Ethiopia

Purpose This study aims to analyze the temperature variability and change for the past 30 years (1990–2019) and the future 60 years (2030s, 2050s and 2070s) in Wolaita Zone and the surroundings, in Southern Ethiopia. Design/methodology/approach The temperature (maximum and minimum) data of the past 30 years (1990–2019) of ten meteorological stations and the future (2021–2080) data of regional climate models (RCMs) under two representative concentration pathways (RCP4.5 and RCP8.5) were used in this study. The accuracy of RCMs in representing observed temperature data was evaluated against mean absolute error, root-mean-square error, percent bias, Nash–Sutcliffe measure of efficiency, index of agreement (d) and coefficient of determination (R2). The temperature variability was analyzed using the coefficient of variation, and the trend was determined using the Mann–Kendall trend and Sen’s slope tests. Findings The results indicate that the past maximum (Tmax) and minimum (Tmin) temperatures showed low variability (CV = 4.3%) with consistently increasing trends. Similarly, Tmax and Tmin are projected to have low variability in the future years, with upward trends. The Tmax and Tmin are projected to deviate by 0.7°C–1.2°C, 1.3°C–2.2°C and 1.5°C–3.2°C by 2030s, 2050s and 2070s, respectively, under RCP4.5 and RCP8.5, from the baseline. Thus, it can be concluded that temperature has low variability in all periods, with consistently increasing trends. The increasing temperature could have been affecting agricultural production systems in Southern Ethiopia. Research limitations/implications This research did not remove the uncertainties of models (inherited errors of models) in future temperature projections. However, this study did not have any limitation. Therefore, individuals or organizations working on agricultural productivity, food security and sustainable development can use the results and recommendations. Practical implications The globe has been warming due to the increasing temperature; as a result, many adaptation and mitigation measures have been suggested globally and nationally (IPCC, 2021). FAO (2017) indicates that the level of vulnerability to the impacts of climate change varies with geographic location, economy and demography; the adaptation measures need to be local. The detailed information on temperature variability and change in the past and future helps to understand the associated negative impacts on agriculture, hydrology, biodiversity, environment and human well-being, among others. Social implications The projected future climate pattern helps the country devise proactive adaptation and mitigation measures for the associated damages at different levels (from local to national). This could improve the resilience of farmers and the country to climate change impacts. This contributes to achieving sustainable development goals (e.g. no poverty, zero hunger and climate action). This is because the agriculture sector in Ethiopia accounts for 80% of employment, 33% of the gross domestic product and 76% of exports (EPRSS, 2023). Originality/value Temperature is one of the major climate elements affecting agricultural production in rain-fed production systems. Despite this, past studies in Southern Ethiopia considered only the past temperature but not the future climate. Thus, generating detailed information about past and future temperatures is very important to take proactive adaptation measures for reducing climate-associated damages in the agriculture sector in Ethiopia.

Breaking the degeneracy in stellar spectral classification from single wide-band images

The spectral energy distribution (SED) of observed stars in wide-field images is crucial for chromatic point spread function (PSF) modelling methods, which use unresolved stars as integrated spectral samples of the PSF across the field of view. This is particularly important for weak gravitational lensing studies, where precise PSF modelling is essential to get accurate shear measurements. Previous research has demonstrated that the SED of stars can be inferred from low-resolution observations using machine-learning classification algorithms. However, a degeneracy exists between the PSF size, which can vary significantly across the field of view, and the spectral type of stars, leading to strong limitations of such methods. We propose a new SED classification method that incorporates stellar spectral information by using a preliminary PSF model, thereby breaking this degeneracy and enhancing the classification accuracy. Our method involves calculating a set of similarity features between an observed star and a preliminary PSF model at different wavelengths and applying a support vector machine to these similarity features to classify the observed star into a specific stellar class. The proposed approach achieves a 91% top-two accuracy, surpassing machine-learning methods that do not consider the spectral variation of the PSF. Additionally, we examined the impact of PSF modelling errors on the spectral classification accuracy.

Kriging data with measurement error: A review and a generalized approach

Filtered kriging with parametric error (FKPE) is a mathematically sound method that generalizes and addresses the oversimplifications of previous kriging algorithms designed to filter error. The proposed approach is developed for handling grade-dependent (heteroscedastic), non-stationary, and spatially correlated sampling errors. The covariance between each pair of measurements or nodes is estimated from their model of errors and the spatial continuity of the underlying process. This research is driven by the fact that sampling and analytical errors are inherent in the samples used in the mining industry. In recent decades, data from quality control programs monitoring these errors have become widely available. FKPE handles more types of errors than other kriging algorithms to filter error and, for any number of subsets, it avoids jointly modelling the models of co-regionalization required by co-kriging methods. The precision gains of FKPE over other methods depend on how well the error model is fit to the data. Therefore, a detailed analysis of the error model and how to estimate its components is discussed. In a five-data toy example with high independent and grade-dependent error, kriging methods with oversimplified error models overweighted a high-grade datum by 50% compared to FKPE estimates. The performance of FKPE and other methods is illustrated by two synthetic examples.

Evaluation of atmospheric circulation of CMIP6 models for extreme temperature events using Latent Dirichlet Allocation

Abstract For climate models to continue improving, we need to uncover as many discrepancies they have with reality as possible. In particular, evaluating the representation of extreme events is important but challenging owing to their rarity. Here, we study how general circulation models reproduce large-scale atmospheric circulation associated with extreme temperature events. To this end, we apply Latent Dirichlet Allocation (LDA), a dimensionality reduction method, to a set of sea-level pressure ERA5 maps over the north-Atlantic region. LDA provides a basis of sparse latent modes called “motifs” that consist of localized objects at synoptic scale. Any pressure map can be approximated by a generally sparse combination of motifs, whose coefficients are called the weights, containing local information about large-scale circulation. Weights statistics can be used to locally characterize circulation patterns, in general and during extreme events, allowing for detailed comparison of datasets. For four CMIP6 models and reanalysis, we quantify local circulation errors and identify model-agnostic and model-specific biases. On average, large-scale circulation is well predicted by all models, but model errors are increased for heatwaves and cold spells. Significant errors were found to be associated with Mediterranean motifs for all models in all cases. In addition, the combination of motif and temperature error can discriminate between models in the general and cold spell cases, while models perform similarly on heatwaves. The sparse characterization provided by LDA analysis is therefore well suited for model preselection for the study of extreme events.

Enhanced Terrain-Referenced Navigation Through Adaptive Radar Altimeter Error Estimation with Monte Carlo Sampling

Abstract Terrain-referenced navigation serves as a reliable backup system for military aviation, but its performance is often compromised by inaccurate radar altimeter noise estimation. The wide beamwidth of radar altimeters causes complex error characteristics that vary with flight conditions and radar–terrain interactions. A mismatch between assumed and actual measurement error levels can lead to degraded navigation accuracy and potential filter divergence. This issue is exacerbated in mission scenarios such as low-altitude penetrations that require frequent rolling and pitching maneuvers. This study proposes a novel radar altimeter error modeling approach that uses Monte Carlo sampling to estimate measurement error levels on-the-fly. By incorporating a radar–terrain interaction measurement model, the method provides tighter error bounds to an extended Kalman filter, improving its response to state- and terrain-dependent error characteristics. Simulation across four distinct scenarios show that the proposed approach reduces the average divergence rate from 2.75 to 0.25%, and decreases horizontal position error by 32.3% in RMSE and 31.7% in CEP.

The Random RFPA Method for Modelling Rock Failure

Abstract The random rock failure process analysis (RRFPA) method was developed in this research to characterize the material spatial variability and uncertainty in rock failure modelling. The random field theory (RFT) was integrated with the traditional rock failure process analysis (RFPA) to model rock heterogeneity. In this approach, the variation of rock properties is represented as a function of relative distance, such that the influence of material intrinsic correlation on its fracturing behaviour can be appropriately captured. To validate the theory, 300 RRFPA simulations were conducted to investigate the failure characteristics of rock samples under compressive loading. The results showed that by incorporating a spectrum of material properties, the numerical outcomes exhibited distinct upper and lower bounds of stress across all testing scenarios, closely aligning with the experimental relationships. The histograms for uniaxial compressive strength and elastic modulus showed that both properties followed normal distributions, with the average values of 10.099 MPa and 1.818 GPa, respectively. The corresponding coefficients of variation were 0.450 and 0.038. The localized failure tended to result in a more rapid release of acoustic emission energy, but generated smaller cumulative energy compared to the overall failure pattern. In general, the maximum relative error of the RRFPA model was only 0.66% for uniaxial compressive strength, elastic modulus, and critical axial strain.

Reliable, efficient, and scalable photonic inverse design empowered by physics-inspired deep learning

Abstract On-chip computing metasystems composed of multilayer metamaterials have the potential to become the next-generation computing hardware endowed with light-speed processing ability and low power consumption but are hindered by current design paradigms. To date, neither numerical nor analytical methods can balance efficiency and accuracy of the design process. To address the issue, a physics-inspired deep learning architecture termed electromagnetic neural network (EMNN) is proposed to enable an efficient, reliable, and flexible paradigm of inverse design. EMNN consists of two parts: EMNN Netlet serves as a local electromagnetic field solver; Huygens–Fresnel Stitch is used for concatenating local predictions. It can make direct, rapid, and accurate predictions of full-wave field based on input fields of arbitrary variations and structures of nonfixed size. With the aid of EMNN, we design computing metasystems that can perform handwritten digit recognition and speech command recognition. EMNN increases the design speed by 17,000 times than that of the analytical model and reduces the modeling error by two orders of magnitude compared to the numerical model. By integrating deep learning techniques with fundamental physical principle, EMNN manifests great interpretability and generalization ability beyond conventional networks. Additionally, it innovates a design paradigm that guarantees both high efficiency and high fidelity. Furthermore, the flexible paradigm can be applicable to the unprecedentedly challenging design of large-scale, high-degree-of-freedom, and functionally complex devices embodied by on-chip optical diffractive networks, so as to further promote the development of computing metasystems.

A hybrid model for carbon price forecasting based on SSA-NSTransformer: Considering the role of multi-stage carbon reduction targets.

In order to reduce the unpredictability of carbon prices caused by their increasingly prominent environmental and market attributes, and to minimize their negative impact on carbon trading, further research on forecasting models for carbon price is urgently needed. To improve the accuracy of prediction, this paper proposes a carbon price forecasting method based on SSA-NSTransformer. The method includes four main steps: Firstly, decomposition of carbon price signals, using Singular Spectrum Analysis to remove noise signals; Secondly, analysis of influencing factors, using Random Forest to identify and select key influencing factors of carbon price signal components from energy price, financial market, socio-economic, and environmental aspects; Furthermore, influencing factors prediction, considering the impact of different carbon reduction targets and predicting future trends of influencing factors; And finally, carbon price prediction, considering the impact of factors based on multi-stage carbon reduction targets, using Non-stationary Transformer to predict the signal components of carbon prices, reconstructing the carbon price time series, and testing the model accuracy. This paper uses historical data of ICE-EUA closing prices and influencing factors as samples, and the model error comparison results confirm that the proposed hybrid model for carbon price forecasting has significant advantages, which can provide scientific and accurate reference for carbon market participants' behaviors.

Study on the influence of square fiber diameter quality on the optical characteristics of lobster eye X-ray micro pore optics

Background The lobster eye micro pore optics (MPO) plays a pivotal role in X-ray focusing, composed of thousands of hollow square microfibers. The channel error in MPO can profoundly impact its focusing performance. Due to the complex manufacturing process of MPO, there are numerous factors that can contribute to channel errors. Objective This paper investigates the impact of two key quality indicators of fiber, i.e., diameter precision and ovality, on the focusing performance of flat MPO. Methods During the actual production process of MPO, fibers with varying diameter precision and ovality are utilized, and point-to-point vacuum X-ray focusing equipment is used to assess MPO's focusing performance. Channel error models related to fiber diameter accuracy and ovality are established in the simulation. Results Experiments show that both the diameter precision and ovality of fiber influence MPO focusing abilities, with diameter precision primarily affecting the intensity and uniformity of the central point focus and the parallelism of the line foci, while ovality mainly affects the intensity and continuity of the line foci. Numerical simulation results reveal that tilt channel errors significantly affect the X-ray focusing effects. Conclusions These findings hold important guiding significance for the preparation process of square fibers and high quality X-ray focusing device.