Prediction Error Research Articles

The illusion of intrinsic meaning: reassessing conscious experience

The illusion of intrinsic meaning in predictive coding through cognitive artifacts to minimize prediction errors points toward a functionalist attempt at understanding conscious experience. It examines how conscious experience serves functional roles in predictive coding and symbolic cognition systems within the brain. By addressing recent developments in diverse fields like cognitive neuroscience, philosophy, and artificial intelligence, we argue that conscious experience emerges from the need to construct coherent narratives for survival and decision-making. Additionally, the paper explores the implications for artificial intelligence, suggesting that artificial systems could develop analogous cognitive artifacts through predictive models without subjective awareness, contributing to a functionalist understanding of consciousness, and further advancing the discussion on the nature of conscious experience in biological and artificial systems.

State estimation of magnetorheological suspension of all-terrain vehicle based on a novel adaptive Sage–Husa Kalman filtering

Abstract For the magnetorheological (MR) suspension control system of all-terrain vehicles (ATVs), state estimation is an effective method to obtain system feedback signals that cannot be directly measured by sensors. However, when confronted with modeling errors and sudden changes in sensor noise during complex road driving, conventional estimation methods with fixed parameters encounter challenges in accurately estimating the states of ATV suspension system. To address this issue, this paper introduces a novel adaptive Sage-Husa Kalman filter (ASHKF) algorithm to estimate the sprung and unsprung velocity of ATV suspension system. The algorithm uses exponential weighting function and gradient detection function to adaptively adjust the attenuation coefficient according to the driving conditions of the ATV, thereby realizing real-time correction of the covariance matrix of the prediction error. Ultimately, through simulation and real-vehicle testing, it is demonstrated that the designed ASHKF is able to effectively improve the state estimation accuracy of the speed signal of the suspension system under off-road driving conditions with low-frequency noise and outlying disturbances, and the accuracy is improved by 62.70% compared with that of the conventional Sage-Husa Kalman filter (SHKF).

A game-theoretic driver steering model with individual risk perception field generation

Driver-automation shared steering control (SSC) has emerged as a promising technology for enhancing vehicle safety, but desire to achieve seamless collaboration between the driver and automation requires an in-depth understanding of driver steering behavior in interaction with automation. In this paper, we introduce a game-theoretic driver steering model with individual risk perception field generation. Firstly, a driver risk perception field is developed based on a novel concept of potential injury risk (PIR) to provide a quantitative estimation of the driver’s perceived driving risk. This approach offers an explicit and physically meaningful structure for simulating the driver’s risk perception process and elucidating the reasons for discrepancies in risk perception. Then, this driver risk perception field is integrated into the framework of non-cooperative Nash game to model the steering interaction between the driver and automation, and the analytical expression of Nash equilibrium is derived in detail. The resulting combined driver model effectively captures the driver adaptation at both the control and planning levels. Next, the key parameters of the combined driver model and its comparators are identified using measured driver steering behavior data from thirty subjects in a series of driving simulator experiments. Finally, the effectiveness and superiority of the combined driver model is validated through a comprehensive comparative analysis. The results demonstrate that the combined driver model achieves the lowest prediction errors compared to its comparators and effectively captures the individual differences in risk perception and steering behavior.

A novel hybrid method for predicting maglev train-induced environmental vibrations combining field measurement, 2.5D FEM modelling and multi-objective optimization

The maglev system has been developed rapidly in recently years since its advantage in mass transportation, ride comfort and energy efficiency, which also caused an increasing concern on the maglev train-induced environmental vibrations. This paper presents a novel hybrid framework for predicting high-speed maglev train-induced environmental vibrations combining field measurement, 2.5D FEM simulation and multi-objective optimization. First, a maglev train-guideway-soil coupled model considering guideway irregularities was proposed to calculate the ground vibration using 2.5D FEM. Then, the guideway irregularity spectrum and soil damping are inverted by the optimization process using the NSGA3 genetic algorithm, in which the differences of measured and calculated vertical vibrations of measuring points are chosen as objective functions. With the inverted guideway irregularity spectrum and soil damping, the ground vibrations were calculated using the 2.5D FEM model and compared with the field measurements. It was found that the inverted PSD spectrum of guideway irregularity generally decreases as the spatial frequency increases, exhibiting a similar trend as the measured ones. The results obtained by the proposed hybrid prediction method agree well with the field measurement in both time domain and frequency domain. The prediction error for the vertical vibration level of all measuring points induced by five maglev train passing are mainly less than 10 %, indicating that the proposed hybrid method provided a reliable and effective method for vibration prediction combining field measurement and numerical simulation.

A fact based analysis of decision trees for improving reliability in cloud computing

The popularity of cloud computing (CC) has increased significantly in recent years due to its cost-effectiveness and simplified resource allocation. Owing to the exponential rise of cloud computing in the past decade, many corporations and businesses have moved to the cloud to ensure accessibility, scalability, and transparency. The proposed research involves comparing the accuracy and fault prediction of five machine learning algorithms: AdaBoostM1, Bagging, Decision Tree (J48), Deep Learning (Dl4jMLP), and Naive Bayes Tree (NB Tree). The results from secondary data analysis indicate that the Central Processing Unit CPU-Mem Multi classifier has the highest accuracy percentage and the least amount of fault prediction. This holds for the Decision Tree (J48) classifier with an accuracy rate of 89.71% for 80/20, 90.28% for 70/30, and 92.82% for 10-fold cross-validation. Additionally, the Hard Disk Drive HDD-Mono classifier has an accuracy rate of 90.35% for 80/20, 92.35% for 70/30, and 90.49% for 10-fold cross-validation. The AdaBoostM1 classifier was found to have the highest accuracy percentage and the least amount of fault prediction for the HDD Multi classifier with an accuracy rate of 93.63% for 80/20, 90.09% for 70/30, and 88.92% for 10-fold cross-validation. Finally, the CPU-Mem Mono classifier has an accuracy rate of 77.87% for 80/20, 77.01% for 70/30, and 77.06% for 10-fold cross-validation. Based on the primary data results, the Naive Bayes Tree (NB Tree) classifier is found to have the highest accuracy rate with less fault prediction of 97.05% for 80/20, 96.09% for 70/30, and 96.78% for 10 folds cross-validation. However, the algorithm complexity is not good, taking 1.01 seconds. On the other hand, the Decision Tree (J48) has the second-highest accuracy rate of 96.78%, 95.95%, and 96.78% for 80/20, 70/30, and 10-fold cross-validation, respectively. J48 also has less fault prediction but with a good algorithm complexity of 0.11 seconds. The difference in accuracy and less fault prediction between NB Tree and J48 is only 0.9%, but the difference in time complexity is 9 seconds. Based on the results, we have decided to make modifications to the Decision Tree (J48) algorithm. This method has been proposed as it offers the highest accuracy and less fault prediction errors, with 97.05% accuracy for the 80/20 split, 96.42% for the 70/30 split, and 97.07% for the 10-fold cross-validation.

Comprehensive propagation of errors for the prediction of woody biomass

Abstract Managing vegetation to sequester carbon in biomass requires estimates to meet standards for accuracy, with methods that are transparent, verifiable and cost‐effective. Allometric models are commonly used to predict biomass from non‐destructive field inventory data. Although a number of studies have addressed biomass error propagation, none have provided a general set of methods for linking errors all the way from initial allometric model development through to the final site‐based biomass prediction, for both above‐ and below‐ground biomass. Error sources in total biomass (above‐ + below‐ground) were quantified using a combination of analytical and Monte Carlo methods, illustrated with four contrasting case studies using either site‐ and‐species‐specific, species‐specific or generalised allometric models. Sampling error was found to be the most important contributor to site‐level biomass uncertainty, arising from the interaction between spatial variability and the field sampling design. The contribution of allometric model covariance to total error was also quantified, with errors in the determination of moisture content during allometric model development identified as a potentially important yet often overlooked error source. Application of different allometric models to the same inventory data suggested the error from generalised models was no greater than that from site‐ or species‐specific models, with increases in the generalised model prediction error balanced by decreases in other error sources associated with the increased sample size on which generalised models are based. Recommendations for reducing errors in predicted biomass include increasing field survey sample size, adopting field survey designs that ensure spatial representativeness and improving moisture content measurement protocols and increasing the moisture content sample size during allometric model development. To reduce costs while maintaining acceptable accuracy, the use of generalised allometric models is recommended, with the caveat that additional biomass sampling for model validation may be required to limit the potential for biased predictions.

A Multimodal Machine Learning Model in Pneumonia Patients Hospital Length of Stay Prediction

Hospital overcrowding, driven by both structural management challenges and widespread medical emergencies, has prompted extensive research into machine learning (ML) solutions for predicting patient length of stay (LOS) to optimize bed allocation. While many existing models simplify the LOS prediction problem to a classification task, predicting broad ranges of hospital days, an exact day-based regression model is often crucial for precise planning. Additionally, available data are typically limited and heterogeneous, often collected from a small patient cohort. To address these challenges, we present a novel multimodal ML framework that combines imaging and clinical data to enhance LOS prediction accuracy. Specifically, our approach uses the following: (i) feature extraction from chest CT scans via a convolutional neural network (CNN), (ii) their integration with clinically relevant tabular data from patient exams, refined through a feature selection system to retain only significant predictors. As a case study, we applied this framework to pneumonia patient data collected during the COVID-19 pandemic at two hospitals in Naples, Italy—one specializing in infectious diseases and the other general-purpose. Under our experimental setup, the proposed system achieved an average prediction error of only three days, demonstrating its potential to improve patient flow management in critical care environments.

Validation of a spatial model of background radiation using personal measurements in children

BackgroundLimited knowledge about the prediction accuracy of exposure models hinders the interpretation of results from epidemiological studies on childhood cancer risks associated with exposure to background gamma-radiation. ObjectiveWe aimed to validate a spatial exposure model that we recently developed for Switzerland. MethodsWe used individual exposure measurements conducted with D-Shuttle dosimeters by 149 children throughout the country. We ran linear regression models fitting the measured exposure against predictions from the newly developed model, and compared results with the predictions from an earlier model. We further used variograms to investigate the spatial correlation of estimation errors. ResultsThe prediction accuracy of the newly developed exposure model was modest (R2 = 0.2), but better than the earlier model (R2 = 0.13). Prediction errors revealed weak spatial correlation. DiscussionAlthough the new exposure model marks an improvement, the modest prediction accuracy and the remaining spatial correlation of errors show room for further improvement. Our study highlights the need for validation of exposure models for background gamma-radiation used in epidemiological studies.

Stock Price Prediction System LSTM Based on Deep Learning

Stock price prediction is a highly complex task due to the inherent volatility and non-linear patterns of financial markets. Recent advancements in deep learning, particularly Long Short-Term Memory (LSTM) networks, have yielded promising results in time series forecasting. This paper provides an in-depth analysis of LSTM models applied to the stock price prediction of major technology companies, including Apple (AAPL), Tesla (TSLA), Microsoft (MSFT), and Nvidia (NVDA). The study utilizes historical stock data to train and evaluate LSTM models, comparing their performance against various other regression models such as Linear Regression, Support Vector Regression (SVR), Decision Tree Regression, and Gradient Boosting Decision Trees (GBDT).The results demonstrate that the LSTM model outperforms traditional models, particularly in its ability to capture long-term dependencies and manage complex, volatile stock data. Among the companies analyzed, the LSTM model achieved the most accurate predictions for Apple, as indicated by the lowest Mean Squared Error (MSE). Tesla's predictions, while less accurate due to high volatility, still showed improvement compared to other models. In conclusion, the LSTM network proved to be the most effective method for stock price prediction in this study, particularly excelling in forecasting long-term trends and reducing prediction errors for volatile stocks.

Quantification of Amlodipine Maleate Content in Amorphous Solid Dispersions Produced by Fluidized Bed Granulation Using Process Analytical Technology Tools

Background: Active pharmaceutical ingredient (API) content is a critical quality attribute (CQA) of amorphous solid dispersions (ASDs) prepared by spraying a solution of APIs and polymers onto the excipients in fluid bed granulator. This study presents four methods for quantifying API content during ASD preparation. Methods: Raman and three near-infrared (NIR) process analysers were utilized to develop methods for API quantification. Four partial least squares (PLS) models were developed using measurements from three granulation batches, with an additional batch used to evaluate model predictability. Models performance was assessed using metrics such as root mean square error of prediction (RMSEP), root mean square error of cross-validation (RMSECV), residual prediction deviation (RPD), and others. Results: Off-line and at-line NIR models were identified as suitable for process control applications. Additionally, at-line Raman measurements effectively predicted the endpoint of the spraying phase. Conclusions: To the best of authors’ knowledge, this is the first study focused on monitoring API content during fluidized bed granulation (FBG) used for ASD preparation. The findings provide novel insights into the application of Raman and NIR process analysers with PLS modelling for monitoring and controlling ASD preparation processes.

Real-time monitoring of multiparameter in 1,3-propanediol production process with near-infrared spectroscopy

1,3-propanediol (1,3-PDO) is a significant product of fermentation, with glycerol serving as the primary substrate in most cases. Bioprocess control based on real-time information of feedstock and main products is crucial for reducing the cost of production. However, rapid quantification of 1,3-PDO and glycerol remains challenging due to their highly similar molecular structures. In this study, the feasibility of near-infrared (NIR) spectroscopy to monitor 1,3-PDO, glycerol, acetate, and butyrate concentrations in the fermentation process using strain Clostridium pasteurianum was evaluated. NIR spectra were acquired through at-line measurement involving sampling and ex-situ analysis or on-line measurement with a fiber optic probe immersed in fermentation broth, integrated with Partial Least Squares (PLS) regression to establish calibration models on a laboratory-scale and pilot-scale. The best PLS regressions of 1,3-PDO, glycerol, acetate, and butyrate with two measurement approaches provided excellent performance, with the root-mean-squared errors of prediction (RMSEP) of 1.656g/L, 1.502g/L, 0.746g/L, and 0.557g/L in at-line measurement and 1.113g/L, 1.581g/L, 0.415g/L, and 0.526g/L in on-line measurement. The cross-scale application performance of at-line measurement was evaluated by an external fermentation trial and an acceptable result was achieved. At-line measurement technique represents a superior choice for the optimization of fermentation process since the robustness across varying fermentation scales and its applicability in multiple bioreactors. Thus, a calibration model developed for one bioreactor is likely to be used in other bioreactors, which enables the reduction of modeling costs. On-line measurement technique, owing to its automated operation and frequent data acquisition, enables real-time monitoring and precise control of the fermentation process, thereby reducing cost and improving production efficiency.

XSCAN: Explainable solder joint defect probability prediction through solder paste printing status with imbalanced data

This research addresses challenges in Surface Mount Technology (SMT) related to solder joint quality prediction, focusing on the initial solder paste printing stage. Recognizing that over 50% of defects originate at the printing stage, this research delves into establishing a direct correlation between printing quality and joint quality. Traditional approaches have limitations in accurately predicting defects due to isolated treatment of printing quality indicators, scarce explainability of prediction models, and lack of joint defect data. This research introduces a novel framework, XSCAN, aimed at predicting the probabilities of solder joint defects from the states of the printed solder paste. This is accomplished by using a generative adversarial network (GAN) to synthesize additional defect data and segment the feature space of printing indicators using customized decision trees to minimize defect probability prediction error. Specifically, XSCAN optimizes generative model structures using decision tree prediction results focused on defects, generating valuable defect information to help feature space partition. Also, pruning rules are designed to handle imbalanced data and improve defect prediction. They enhance explainability by defining safe and high-risk zones for solder paste quality. XSCAN outperforms all other baselines when tested on real-world datasets of chip resistors. It achieves the lowest prediction error and provides different warning levels for potential joint defects. XSCAN takes a proactive approach to improve manufacturing quality while addressing data imbalance and model explainability challenges. It provides practical insights to enhance SMT processes and reduce waste and rework costs.

Ecological momentary assessment models trajectories of expectancy following exposure: A proof-of-concept pilot study

Background and objectivesThe Inhibitory Retrieval Approach to exposure therapy for fears and anxiety emphasizes prediction error as one of several strategies for improving outcomes. Prediction error depends on disconfirmation of expectancies for the feared outcome, and thus exposure strategies that derive from inhibitory retrieval approaches emphasize expectancy violation during exposure. However, research studies examining expectancy violation in exposure therapy have treated expectancy as a stable characteristic, assuming that expectancy following an exposure exercise remains constant over time. This brief report outlines two different uses of a methodology for using ecological momentary assessment (EMA) to assess between-session expectancy following exposure during treatment for anxiety, and reports on pilot trial results. MethodsAdults with social anxiety disorder (N = 12) and spider phobia (N = 31) taking part in larger trials investigating exposure therapy completed EMA questionnaires assessing expectancy for their feared outcome for 2–4 days following each of two exposure sessions. ResultsExpectancy ratings decreased from pre-to post-exposure and remained stable for 2–4 days following exposure. LimitationsThis pilot study used a very limited sample size and should be replicated in a larger sample. ConclusionsExpectancy for feared outcome may be assessed using EMA following exposure sessions. Pilot results suggest that expectancy decreases immediately following exposures and remains stable afterwards.

An AI model for predicting the spatiotemporal evolution process of coastal waves by using the Improved-STID algorithm

Accurate and rapid prediction of water wave evolution is crucial for ensuring the safe and healthy operation of ocean and coastal engineering. Using hourly wave data from the coastal areas of the Hawaiian Islands from 2019 to 2021, this study developed an artificial intelligence model for rapidly predicting the spatiotemporal evolution of coastal waves by improving the STID algorithm. In the study, the SWAN wave model was used to generate an initial training dataset for driving Improved-STID algorithm, where the Improved-STID algorithm and SWAN model share the same unstructured computational grid with good adaptability to complex topography. The proposed Improved-STID algorithm takes into account the overall spatiotemporal correlations of wave evolution across different regions. The results show that on the basis of prediction efficiency on a par with the original STID, the Improved-STID can be more accurately applied to the spatiotemporal evolution process of coastal waves. The prediction errors in this study are: the RMSE of 0.23 m, MAE of 0.14 m, MAPE of 17.96 %, and the calculation time of the proposed AI model is only about 1.8 % of that of the SWAN wave model.

How to describe the time-dependent dissolution of engineered nanomaterials?

Numerous processes such as solubility, agglomeration/aggregation, or protein corona formation may change over time and significantly affect engineered nanomaterial (ENM) structure, property, and availability, resulting in their reduced or increased toxicological activity. Therefore, understanding the dynamics of these processes is essential for assessing and managing the risks of ENMs during their lifecycle, ensuring safety by design. Of these processes, the importance of solubility (i.e., the ability to release ions from the surface) is undeniable. Thus, we propose a practical approach, the Kalapus equation (KEq), to determine ENMs' dissolution over time. As a proof-of-concept, the KEq was applied to determine the solubility of six commercially used metal and metal oxide nanoparticles over time. The KEq exhibited a higher coefficient of determination (R2=0.995-0.999) than the logarithmic equation (R2=0.835-0.986), and the pseudo-first-order equation (R2=0.915-0.994) over a range of experimental data. The newly introduced Kalapus equation outperformed the logarithmic and pseudo-first-order equations when extrapolating beyond the time range in which solubility was experimentally determined. The mean absolute error in solubility prediction for the KEq was 3.29% and 4.28% for the first and second data points, respectively, significantly lower than the 13.46% and 18.05% observed for the pseudo-first-order/first-order equation. The proposed equation can be used as a part of New Generation Risk Assessment (NGRA) methodology, especially new Integrated Approaches to Testing and Assessments (IATAs).

Predictive Current Error Compensation Based Strong Robust Model Predictive Control for PMSM Drive Systems

Predictive Current Error Compensation Based Strong Robust Model Predictive Control for PMSM Drive Systems

Research on deformation prediction of CFETR multi-purpose overloaded robot based on real-time structural simulator

The CFETR Multi-Purpose Overload Robot (CMOR) is a key subsystem of the China Fusion Engineering Test Reactor (CFETR), which can perform maintenance tasks of the internal components. However, the large slenderness ratio of the structure results in low control accuracy of the CMOR end-effector. This paper proposes a CMOR deformation prediction method based on the real-time structural simulator. The overall deformation model of CMOR is analyzed based on the single joint deformation mechanism. Based on the principle of layered control, the control framework of the CMOR structural simulator is constructed, and the deformation data of CMOR in different positions are calculated based on the finite element method. A hybrid neural network containing a multilayer perceptron, transformer, and attention mechanism is designed to train the CMOR deformation prediction model. The training results show that the deformation prediction model converges quickly and fits the deformation of the CMOR structure well with small prediction errors. Finally, the real-time structure simulator is developed based on the deformation prediction model, and the deformation of CMOR is reconstructed with the update frequency of 2 Hz and the absolute error at any point within ±10 mm, which verifies the correctness of the CMOR structural deformation prediction method.

Efficient and accurate determination of crack propagation rate without unloading-induced crack closure

The accurate delineation of the fatigue crack growth rate (da/dN) model, encompassing the short crack stage, assumes paramount importance in the operation and maintenance of transportation vehicles such as aircraft. Nonetheless, the determination of the da/dN encounters challenges, notably crack closure engendered by unloading observations on the specimen and the attendant issue of exorbitant test expenditures due to low load amplitude (ΔP), thereby impeding the precise establishment of the da/dN model, particularly in the short crack stage. Hence, this study initially posits a novel crack observation methodology employing a self-engineered stress transfer apparatus to sustain the stress (σ) on the specimen at its closure stress (σcl) level post-unloading, thus mitigating the impact of crack closure on crack observation and augmenting the accuracy of da/dN curve determination. Subsequently, building upon this innovation, a fatigue crack propagation (FCP) test protocol tailored for the short crack stage is advanced, principally by substantially augmenting ΔP to surmount the cost constraints of the test, thereby broadening the scope of da/dN curve determination to 10−8 mm/cycle. Lastly, premised on the acquisition of high precision and comprehensive da/dN curves, a da/dN model is posited by integrating with the extant da/dN model, which can comprehensively and precisely delineate the FCP behavior across the entire crack growth spectrum, particularly in the short crack stage. Illustratively demonstrated through a case study on 42CrMo steel crankshaft, a fatigue life (Nf) prediction investigation conducted based on this da/dN model substantiates its efficacy in accurately assessing the FCP behavior of crankshaft materials vis-à-vis existing models, thereby reducing the Nf prediction error by 14.64 %.

Distinguishing expectation and attention effects in processing temporal patterns of visual input

The current study investigated how the brain sets up expectations from stimulus regularities by evaluating the neural responses to expectations driven implicitly (by the stimuli themselves) and explicitly (by task demands). How the brain uses prior information to create expectations and what role attention plays in forming or holding predictions to efficiently respond to incoming sensory information is still debated. We presented temporal patterns of visual input while recording EEG under two different task conditions. When the patterns were task-relevant and pattern recognition was required to perform the button press task, three different event-related brain potentials (ERPs) were elicited, each reflecting a different aspect of pattern expectation. In contrast, when the patterns were task-irrelevant, none of the neural indicators of pattern recognition or pattern violation detection were observed to the same temporally structured sequences. Thus, results revealed a clear distinction between expectation and attention that was prompted by task requirements. These results provide complementary pieces of evidence that implicit exposure to a stimulus pattern may not be sufficient to drive neural effects of expectations that lead to predictive error responses. Task-driven attentional control can dissociate from stimulus-driven expectations, to effectively minimize distracting information and maximize attentional regulation.

A recommendation model for optimizing transfer learning hyper-parameter settings in building heat load prediction with limited data samples

The transfer learning method has gained increasing attention in the domain of building load prediction, particularly in scenarios with limited data samples. Its core principle involves leveraging knowledge obtained from abundant data in source buildings to aid the learning process of models for the target buildings. Existing research has predominantly concentrated on optimizing the selection of source building data to improve transfer learning effectiveness, while the optimization of transfer learning hyper-parameter settings is often neglected. This study proposes a recommendation model tailored for transfer learning hyper-parameter settings in the context of small sample prediction for building heat loads. The objective is to automatically suggest suitable transfer learning hyper-parameter combination based on the specific features of the building heat load data samples. In this study, 200 real building profiles were utilized to generate the input–output dataset required for the recommendation model. By employing data mining techniques such as clustering and classification, the correlation between the features of source building data and the most effective transfer learning hyper-parameter combination is investigated. The developed recommendation model for optimal transfer learning hyper-parameter settings achieves a classification accuracy of 90.5%，and the performance evaluation was conducted using an additional dataset of 30 source buildings. The results show that by employing this recommendation model, the prediction error of the target buildings can be reduced by 0.12% to 6.64% compared to the conventional method of empirically determining transfer learning hyper-parameter settings.