Loss Of Accuracy Research Articles

Implementation of Distributed Machine Learning in Cloud Calculation: Evidence from Consumer Behavior Prediction

Abstract. In the field of machine learning, distributed machine learning (DML) has gained massive popularity in research and appeared in a wide range of applications including transportation, medicine, and business. Distributed machine learning showed huge potential in processing large amounts of data while keeping the data private. In this paper, a distributed implementation of a deep neural network (DNN) with a large dataset from a telecom company, aiming to predict consumer behavior based on usage data is carried out. Various datasets with distinct heterogeneity settings were applied to the network, simulating a real-world cloud application scenario. Comparisons between both the centralized model and the distributed models were made to analyze the impact on model performance with device heterogeneity and data heterogeneity. These result indicate that under certain circumstances, DML models showed better accuracy and lower loss compared to the centralized model. The paper provided some insights and made some assumptions on how the data and the parameters affect the model performance.

Discrete Element Modeling of Irregular-Shaped Particles Using an Automatic Generation Algorithm-Based Multi-Sphere Model

Within the Discrete Element Method (DEM) framework, the multi-sphere model capable of creating irregular-shaped particles by several sub-spheres attracts increasing attention. Traditionally, the creation of irregular-shaped particles using the multi-sphere model requires manual determination of the position and diameter of each sub-sphere, which extremely increases the difficulty and workload of DEM modeling of irregular-shaped particles. To tackle this issue, a new automatic generation algorithm that helps to create arbitrary-shaped multi-sphere particles automatically was developed in this study. The quality of created multi-sphere particles can be easily controlled by adjusting the number of sub-spheres and size factor referred to as the size ratio of the smallest sub-sphere to the largest sub-sphere. To verify the effectiveness of the newly developed automatic generation algorithm, two verification tests were conducted in this study: (i) The packing behavior of irregular-shaped particles in a rectangular container and (ii) the mechanical behavior of irregular-shaped particles under the uniaxial compression. The fill heights of the simulations were in line with those of the experiments with an error of less than 8%. The mechanical behavior of irregular-shaped particles under the uniaxial compression can also be well reproduced. Besides, the influence of important parameters on simulation accuracy was also investigated. The results demonstrated that the newly developed automatic generation algorithm can accurately create complex-shaped particles with a small number of sub-spheres. Increasing the number of sub-spheres will improve the simulation accuracy but at the expense of computational efficiency. A larger size factor leads to higher computational efficiency in DEM simulation but is accompanied by a significant loss of accuracy.

TEC-CNN: Towards Efficient Compressing Convolutional Neural Nets with Low-rank Tensor Decomposition

Most state-of-the-art convolutional neural networks (CNNs) are characterised by excessive parameterisation, leading to a high computational burden. Tensor decomposition has emerged as a model reduction technique for compressing deep neural networks. Previous approaches have predominantly relied on either Tucker decomposition or Canonical Polyadic (CP) decomposition for CNNs. However, CP decomposition exhibits exceptional compression capabilities in comparison to Tucker decomposition, which results in a more pronounced accuracy loss. This paper introduces an efficient model compression method, termed TEC-CNN, designed to achieve significant compression while preserving accuracy levels comparable to those of the original models. In TEC-CNN, convolutional layers are identified to obtain convolutional kernels by analysing given models under the principles of low-rank tensor decomposition, and then, calculating the ranks of convolutional kernels. Furthermore, an efficient decomposition schema for the convolutional kernel is proposed with approximate kernel tensor for reducing parameters. Additionally, a novel format of a convolutional sequence is presented and constructed with a reduced number of parameters to replace the original convolutional layers. Finally, the effectiveness of TEC-CNN is assessed across a range of computer vision tasks. For instance, in CIFAR-100 classification, ResNet18 is compressed to 4.1 MB, while Unext, when applied to image segmentation using the International Skin Imaging Collaboration (ISIC) dataset, is reduced to 3.419 MB. When employed for fire object detection with Yolov7, TEC-CNN achieves a model size reduction of 71.6 MB. Comprehensive experimental results underscore that our approach achieves significant model compression while preserving model performance.

CAPP: Context‐Aware Privacy‐Preserving Continuous Authentication in Smartphones

ABSTRACTThe pervasive use of smart devices in everyday life has increased the risk of theft and cyberattacks, necessitating robust and continuous behavioral authentication techniques during every interactive session. Traditional authentication methods are insufficient to address these evolving threats, as they often fail to maintain security throughout the session. To address this challenge, we propose a continuous authentication scheme that leverages touch and motion sensors, incorporating user context to enhance security. However, building an effective authentication model requires creating a behavioral template by aggregating data from multiple users. Moreover, outsourcing sensitive sensor data to third‐party servers for processing introduces the risk of privacy breaches, such as behavioral profiling or the theft of personal information. To mitigate these risks, we designed CAPP—a context‐aware privacy‐preserving data outsourcing scheme for continuous authentication during live sessions. CAPP employs Format Preserving Encryption (FPE) to encrypt the sensor data, ensuring that the data's format remains intact, which allows the machine learning model to function with minimal loss of accuracy. Our solution enables secure and efficient authentication by creating a model on the server side using data from motion sensors like the accelerometer, gyroscope, and magnetometer. Our experimental results demonstrate the effectiveness of the proposed scheme, achieving an accuracy of 87.57% and an Equal Error Rate (EER) of 12.15%, while requiring only 1.69 ms to authenticate a user. These results show that CAPP preserves user privacy and provides a rapid and reliable authentication process, outperforming existing state‐of‐the‐art methods.

Rapid Prediction and Parameter Evaluation of Process-Induced Deformation in L-Shape Structures Based on Feature Selection and Artificial Neural Networks

The process-induced deformation (PID) during the manufacturing of thermosetting composite materials can significantly compromise manufacturing precision. This paper introduces an innovative method that combines a finite element analysis (FEA), feature classification algorithms, and an Artificial Neural Network (ANN) framework to rapidly predict the PID of a typical L-shaped structure. Initially, a comprehensive range of parameters that influence PID are compiled in this research, followed by the generation of a dataset through FEA considering viscoelastic constitutive models, validated by experimental results. Influential parameters are classified using Random Forest and LASSO regression methods, with each parameter rated according to its impact on PID, delineating their varying degrees of importance. Subsequently, through a hyperparameter analysis, an ANN framework is developed to rapidly predict the PID, while also refining the assessment of the parameters’ significance. This innovative approach achieves a computational time reduction of 98% with less than a 5% loss in accuracy, and highlights that under limited computational conditions, considering only a subset or all of the parameters—the peak temperature, corner angle, coefficient of chemical shrinkage, coefficient of thermal expansion, curing pressure, and E1—minimizes accuracy loss. The study demonstrates that machine learning algorithms can effectively address the challenge of predicting composite material PID, providing valuable insights for practical manufacturing applications.

Analytical interference of Burosumab therapy on intact fibroblast growth factor 23 (iFGF23) measurements using an immunoassay: preliminary evaluation

ABSTRACT Our study evaluated the possible interference of Burosumab (human recombinant monoclonal antibody directed against N-terminal domain of FGF23) on the immunoassay of intact FGF23 (iFGF23) with the Liaison XL. The analytical method uses three different antibodies, one of which directed against the N-terminal portion of FGF23. The evaluation of the method accuracy involved the fully automated execution of a dilution test on EDTA plasma from 5 subjects who had not received any monoclonal antibody (mAb), 20 EDTA plasma from patients treated with Burosumab, and 2 EDTA plasma from subjects who had not received any mAb in witch an adequate volume of Burosumab had been added in vitro. One sample with specific diluent (LIAISON® FGF 23) with an adequate volume of Burosumab had been added in vitro. The dilution assay provided highly inaccurate iFGF23 results in samples with therapeutic concentrations of Burosumab and in samples with concentrations below the LoQ (6.5 pg/mL). The addition of Burosumab to the diluent did not produce any analytical interference. Dissociation of iFGF23 from the mAb-target complex in diluted sample could explain the loss of accuracy in the iFGF23 immunoassay using the Liaison XL analyzer. Burosumab could be an interferent in immunoassay procedures of iFGF23.

Compositional reservoir simulation with a high-resolution compact stencil adaptive implicit method

The adaptive implicit method (AIM) is the mainstream approach for compositional reservoir simulation. In its standard form, AIM uses single-point upwinding to reconstruct the fluxes across the interfaces of the control volume, and this leads to substantial numerical diffusion and loss of accuracy. Previous efforts to improve the accuracy of AIM focused on using high-order schemes to reconstruct the interfacial fluxes; those schemes introduced additional numerical nonlinearities to the system of equations or compromised the accuracy of the Jacobian by neglecting the high-order terms in its construction. In this work, we describe a high-resolution compact-stencil (HRCS) AIM. The new scheme is applied to compositional reservoir simulation. In addition to the mixed implicit/explicit time discretization of standard AIM, the HRCS AIM scheme uses a mixed time and space discretization. Specifically, we blend low- and high-order fluxes according to a well defined rule that uses a high-order reconstruction in the explicit regions of the domain and a low-order reconstruction in the implicit regions. This strategy ensures that additional nonlinearities introduced by the high-order reconstruction do not impact the Jacobian matrix, thus preserving the algebraic structure of standard AIM. The HRCS AIM method is demonstrated using several compositional problems. The results indicate substantial gains in accuracy with a small additional computational cost compared with standard AIM. Additionally, HRCS AIM is more robust and has a smaller computational cost compared with its full high-resolution counterpart.

Computer model coupling hemodynamics and oxygen transport in the coronary capillary network: Pulsatile vs. non-pulsatile analysis

Background and Objective:Oxygen transport in the heart is crucial, and its impairment can lead to pathological conditions such as hypoxia, ischemia, and heart failure. However, investigating oxygen transport in the heart using in vivo measurements is difficult due to the small size of the coronary capillaries and their deep embedding within the heart wall. Methods:In this study, we developed a novel computational modeling framework that integrates a 0-D hemodynamic model with a 1-D mass transport model to simulate oxygen transport in/across the coronary capillary network. Results:The model predictions agree with analytical solutions and experimental measurements. The framework is used to simulate the effects of pulsatile vs. non-pulsatile behavior of the capillary hemodynamics on oxygen-related metrics such as the myocardial oxygen consumption (MVO2) and oxygen extraction ratio (OER). Compared to simulations that consider (physiological) pulsatile behaviors of the capillary hemodynamics, the OER is underestimated by less than 9% and the MVO2 is overestimated by less than 5% when the pulsatile behaviors are ignored in the simulations. Statistical analyses show that model predictions of oxygen-related quantities and spatial distribution of oxygen without consideration of the pulsatile behaviors do not significantly differ from those that considered such behaviors (p-values >0.05). Conclusions:This finding provides the basis for reducing the model complexity by ignoring the pulsatility of coronary capillary hemodynamics in the computational framework without a substantial loss of accuracy when predicting oxygen-related metrics.

Assessing the influence of nonlinear mooring restoring forces in the optimization and design of FOWT

Abstract The design of mooring systems plays a pivotal role in the success of Floating Offshore Wind Turbine (FOWT) projects. This process naturally involves iterations, as design and analysis are intricately linked. Despite the inherent nonlinearity of mooring system restoring forces, many designers opt for employing an equivalent linear mooring system stifness matrix to expedite the optimization process in early design stages. This article aims to underscore the limitations associated with relying on the equivalent linear model when compared to accounting for nonlinear restoring forces during optimization processes. To illustrate this point, a case study was conducted using the reference semi-submersible platform VolturnUS-S. The study considered intermediate to large water depths and addressed three different mooring confgurations: catenary, semi-taut, and taut lines. The analysis focused on several critical aspects, including ofset watch circles, line tensions (including those at the anchors), and a cost estimate based on the different models. The findings indicate that, for catenary-based mooring systems, the linear model remains a good approximation, leading to no significant loss of accuracy in the context of early design stages. As the lines become tauter, however, the nonlinearities become more pronounced and the errors involved in the linear model can reach unacceptable levels.

Convective heat transfer parameters in the film cooling with thermal radiation

Thermal radiation is a large contributor to the total heat transfer of combustion systems so that it cannot be neglected, with the improvement of gas temperature and infrared stealth requirements. This paper deeply investigated the convective parameters in film cooling with thermal radiation through theoretical analysis and numerical simulation. A method to obtain the convective driving temperature under radiation was proposed and demonstrated. Research found that the film cooling effectiveness with radiation should be expressed by the mixing temperature rather than the adiabatic wall temperature. Although heat transfer and mass transfer under radiation cannot be theoretically analogized due to the radiant source term in the energy equation, the concentration cooling effectiveness without and with radiation remains approximately the same. The numerical results indicate that although conjugate coupling plays a crucial role in determining convective and radiative heat fluxes, the coupling of radiation and convective parameters is very weak. Radiation has little influence on the film cooling effectiveness (η) and the convective heat transfer coefficient (hf). Convective parameters can be decoupled from radiation with negligible loss of accuracy. This allows the non-radiation convective parameters η and hf to be directly used for calculating the convective heat flux under radiation, with relative deviations of less than 5%.

A deconstruction of methods to derive one-point lensing statistics

Gravitational lensing is a crucial tool for exploring cosmic phenomena, providing insights into galaxy clustering, dark matter, and dark energy. Given the substantial computational demands of N-body simulations, approximate methods like 𝙿𝙸𝙽𝙾𝙲𝙲𝙷𝙸𝙾 and 𝚝𝚞𝚛𝚋𝚘𝙶𝙻 have been proposed as viable alternatives for simulating lensing probability density functions (PDFs). This paper evaluates these methods and their effectiveness across both weak and strong lensing regimes, with a focus in the context where baryonic effects are negligible. Our comparative analysis reveals that these methods are effective for applications where lensing is mild, such as the majority of sources of electromagnetic and gravitational waves. However, both 𝙿𝙸𝙽𝙾𝙲𝙲𝙷𝙸𝙾 and 𝚝𝚞𝚛𝚋𝚘𝙶𝙻 break down for large values of convergence and magnification due to their loss of accuracy in capturing small-scale nonlinear matter fields, owing to oversimplified assumptions about internal halo structures and reliance on perturbation theory. 𝙿𝙸𝙽𝙾𝙲𝙲𝙷𝙸𝙾 yields second-to-fourth moments of the lensing PDFs, which are 6–10% smaller than those resulting from N-body simulations in regimes where baryonic effects are minimal. These findings aim to inform future studies on gravitational lensing of point sources, which are increasingly relevant with upcoming supernova and gravitational wave datasets.

Hierarchical Mixed-Precision Post-Training Quantization for SAR Ship Detection Networks

Convolutional neural network (CNN)-based synthetic aperture radar (SAR) ship detection models operating directly on satellites can reduce transmission latency and improve real-time surveillance capabilities. However, limited satellite platform resources present a significant challenge. Post-training quantization (PTQ) provides an efficient method for pre-training neural networks to effectively reduce memory and computational resources without retraining. Despite this, PTQ faces the challenge of maintaining model accuracy, especially at low-bit quantization (e.g., 4-bit or 2-bit). To address this challenge, we propose a hierarchical mixed-precision post-training quantization (HMPTQ) method for SAR ship detection neural networks to reduce quantization error. This method encompasses a layerwise precision configuration based on reconstruction error and an intra-layer mixed-precision quantization strategy. Specifically, our approach initially utilizes the activation reconstruction error of each layer to gauge the sensitivity necessary for bit allocation, considering the interdependencies among layers, which effectively reduces the complexity of computational sensitivity and achieves more precise quantization allocation. Subsequently, to minimize the quantization error of the layers, an intra-layer mixed-precision quantization strategy based on probability density assigns a greater number of quantization bits to regions where the probability density is low for higher values. Our evaluation on the SSDD, HRSID, and LS-SSDD-v1.0 SAR Ship datasets, using different detection CNN models, shows that the YOLOV9c model with mixed-precision quantization at 4-bit and 2-bit for weights and activations achieves only a 0.28% accuracy loss on the SSDD dataset, while reducing the model size by approximately 80%. Compared to state-of-the-art methods, our approach maintains competitive accuracy, confirming the superior performance of the HMPTQ method over existing quantization techniques.

Deep Learning Ensemble for Flood Probability Analysis

Predicting flood events is complex due to uncertainties from limited gauge data, high data and computational demands of traditional physical models, and challenges in spatial and temporal scaling. This research innovatively uses only three remotely sensed and computed factors: rainfall, runoff and temperature. We also employ three deep learning models—Feedforward Neural Network (FNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM)—along with a deep neural network ensemble (DNNE) using synthetic data to predict future flood probabilities, utilizing the Savitzky–Golay filter for smoothing. Using a hydrometeorological dataset from 1993–2022 for the Nile River basin, six flood predictors were derived. The FNN and LSTM models exhibited high accuracy and stable loss, indicating minimal overfitting, while the CNN showed slight overfitting. Performance metrics revealed that FNN achieved 99.63% accuracy and 0.999886 ROC AUC, CNN had 95.42% accuracy and 0.893218 ROC AUC, and LSTM excelled with 99.82% accuracy and 0.999967 ROC AUC. The DNNE outperformed individual models in reliability and consistency. Runoff and rainfall were the most influential predictors, while temperature had minimal impact.

Integration of drone and machine learning technology for predicting power infrastructure faults efficiently

Power transmission and distribution networks frequently face issues, especially in harsh environments, leading to high maintenance costs and the need for uninterrupted electricity. Current field inspections by skilled personnel are labor-intensive, costly, and slow, often lacking efficiency and posing safety risks. While automated helicopters, flying robots, and climbing robots have been explored for visual inspections, the widespread adoption of automatic vision-based inspection remains limited due to high accuracy demands and unique challenges. This highlights the need for a fully autonomous vision-based system to inspect electrical power infrastructure and predict potential future faults. This research introduces Unmanned Aerial Vehicles (UAVs) as a promising solution for infrastructure inspection, deep learning for data analysis and prediction, and a mathematical model to ensure system accuracy doesn't rely solely on the dataset. As a proof of concept, YOLO V8 was employed to predict electrical faults in insulators, achieving a box loss of 0.525, a classification loss of 0.3887, and a precision of 0.8976, demonstrating high accuracy and low loss.

Модель та метод навчання детектора об’єктів на аерозображенні з оптимізацією робастності та кількості обчислень

The subject of research is Neural network-based object detectors, which are widely used for video image analysis. An increasing number of tasks now demand data processing directly at the source, which limits the available computational resources. However, the vulnerability of neural networks to noise, adversarial attacks, and weight error injections significantly diminishes their robustness and overall effectiveness. The relevant task is to develop models that provide both computational efficiency and robustness against perturbations. This paper investigates a model and method for enhancing the robustness of neural network detectors under limited resources. The objective is to design a model that allocates resources optimally while maintaining stability. To achieve this, the study employs techniques such as dynamic neural networks, robustness optimization, and resilience strategies. The following results were obtained. A detector with a feature extractor based on ViT-S/16, modified with gate modules for dynamic examination was developed. The model was trained on the RSOD dataset and meta-learned on the adaptation results to various perturbations. The model's resistance to random bit inversions in weights (10 % of weights) and to adversarial attacks with perturbation amplitudes up to 3/255 (L∞ norm) was tested. Conclusion. The proposed detector model incorporating dynamic examination and optimized robustness, reduced floating-point operations by over 20 % without loss of accuracy. A novel method for training the detector was developed, combining the RetinaNet loss function with the loss function of gate blocks and applying meta-learning on the adaptation results for various types of synthetic perturbations. Testing demonstrated an increase in accuracy by 11.9 % under the influence of error injection and by 13.2 % under the influence of adversarial attacks.

Order-difference of normalized square envelope spectrum and its applications in early chatter detection of roll grinder

Roll grinder is widely demanded in grinding the surface of rolls that will be used to mill steel strips for high-quality automobile outer panels. The grinding chatter generally leads to poor machining accuracy and massive economic losses. The significance of roll grinder vibration monitoring is to detect the early grinding chatter. However, in harsh working conditions, some chatter detection methods inevitably confuse the fault components and interference components in vibration signals and the chatter feature information cannot be extracted exactly. The dynamical model of roll grinder system reveals that the spectrums of vibration signals contain intrinsic frequency components and chatter frequency components under health and chatter conditions. Based on the analytical modelling, a new health index order-difference of normalized square envelope spectrum (ODNSES) is proposed to detect early chatter and evaluate the health degradation of the roll grinder. The normalized square envelope spectrum (SES) is used as the pretreatment of ODNSES to compare the frequency components in vibration signals under health and early chatter conditions. Furthermore, order-difference of normalized SES is proposed to suppress the noise interference and to enable ODNSES to characterize the grinding chatter process. The detailed simulation research is performed to verify three important properties of ODNSES. The experimental results show that ODNSES would not be interfered by background noise significantly and it is on average 29.12% more sensitivity to impulse amplitudes than classical health indexes. The testing results on actual roll grinder exhibit that ODNSES can detect the early chatter and quantify the abnormal vibration in the grinding process.

Joint optimal PMU and DULR placement in distribution network considering fault reconstruction and state estimation accuracy

Due to the rapid development of the distribution network, there is an urgent need for high-precision state estimation and safe and stable operation, this paper proposes a joint optimal placement method for Phasor Measurement Unit (PMU) and Dual Use Line Relay (DULR) in distribution networks considering fault reconstruction and state estimation accuracy. The proposed method aims to minimize the cost of PMU and DULR configuration. Based on the known pseudo-measurements and Supervisory Control and Data Acquisition (SCADA) measurements, we use the E-optimal standard design of the state estimation error covariance matrix to form the state estimation accuracy index and take the relay protection function of Feeder Terminal Unit (FTU) device and DULR device into consideration of the fault reconstruction. By solving the mixed integer semidefinite programming (MISDP) model, the optimal placement result is obtained, so that the network under normal condition and the network after multiple fault reconstruction can meet the constraints of state estimation accuracy and load loss rate respectively. The proposed method is tested on IEEE 33-node, 123-node systems as an example. The experimental results show that the proposed method is effective and practical.Note to Practitioners—This study is stimulated by the need for optimizing the placement of Micro-PMU (µPMU) in distribution network. Since the new µPMU device DULR has the same relay protection function as FTU, and distributed generation can supply power to the island after failure, this paper adds the load loss rate to the constraint conditions to study the influence of fault reconstruction on the placement scheme. A new two-step SE method considering mixed measurements was used in this paper to improve the computational efficiency of real-time SE of large-scale system while ensuring the accuracy of the calculation results. We have conducted tests in IEEE-33 and IEEE-123 systems. After preliminary verification, compared with the single fault scenario, we find that the placement scheme considering multiple fault scenarios can effectively improve the accuracy of state estimation and reduce the load loss rate of each fault scenario and only increase the cost a little. The increase in the number of DG and contact lines in the network will also effectively improve the feasibility of the placement scheme. The increase of the number of FTU configurations can effectively reduce the configuration of DULR by adding a small amount of µPMU, reduce the total cost as a result. In future studies, we will further consider the integrated configuration of µPMU, DULR and FTU, to meet more failure scenarios while reducing costs.

Optimizing multi-environment trials in the Southern US Rice belt via smart-climate-soil prediction-based models and economic importance.

Rice breeding programs globally have worked to release increasingly productive and climate-smart cultivars, but the genetic gains have been limited for some reasons. One is the capacity for field phenotyping, which presents elevated costs and an unclear approach to defining the number and allocation of multi-environmental trials (MET). To address this challenge, we used soil information and ten years of historical weather data from the USA rice belt, which was translated into rice response based on the rice cardinal temperatures and crop stages. Next, we eliminated those highly correlated Environmental Covariates (ECs) (>0.95) and applied a supervised algorithm for feature selection using two years of data (2021-22) and 25 genotypes evaluated for grain yield in 18 representative locations in the Southern USA. To test the trials' optimization, we performed the joint analysis using prediction-based models in four different scenarios: i) considering trials as non-related, ii) including the environmental relationship matrix calculated from ECs, iii) within clusters; iv) sampling one location per cluster. Finally, we weigh the trial's allocation considering the counties' economic importance and the environmental group to which they belong. Our findings show that eight ECs explained 58% of grain yield variation across sites and 53% of the observed genotype-by-environment interaction. Moreover, it is possible to reduce 28% the number of locations without significant loss in accuracy. Furthermore, the US Rice belt comprises four clusters, with economic importance varying from 13 to 45%. These results will help us better allocate trials in advance and reduce costs without penalizing accuracy.

Pediatric Orally Disintegrating Tablets (ODTs) with Enhanced Palatability Based on Propranolol HCl Coground with Hydroxypropyl-β-Cyclodextrin

Background: Propranolol, largely prescribed as an antihypertensive and antiarrhythmic drug in pediatrics, is characterized by a bitter taste and an astringent aftertaste. Currently, the therapy requires crushing of tablets for adults and their dispersion in water many times a day, leading to loss of dosing accuracy, low palatability, and poor compliance for both patients and caregivers. Objectives: This work aimed to exploit cyclodextrin complexation by cogrinding to develop orally disintegrating tablets (ODTs) endowed with reliable dosing accuracy, good palatability and safety, ease of swallowability, and ultimately better compliance for both pediatric patients and caregivers. Results: Different formulation variables and process parameters were evaluated in preparing ODTs. The technological and morphological characterization and disintegration tests were performed according to official and alternative tests to select the ODT formulation based on the drug Hydroxypropyl-β-cyclodextrin (HPβCD) coground complex form containing Pearlitol® Flash as the diluent and 8% Explotab® as the superdisintegrant, which demonstrated the highest % drug dissolution in simulated saliva and acceptable in vitro palatability assessed by the electronic tongue, confirming the good taste-masking power of HPβCD towards propranolol. Conclusions: Such a new dosage form of propranolol could represent a valid alternative to the common extemporaneous preparations, overcoming the lack of solid formulations of propranolol intended for pediatric use.

Chronobridge: a novel framework for enhanced temporal and relational reasoning in temporal knowledge graphs

The task of predicting entities and relations in Temporal Knowledge Graph (TKG) extrapolation is crucial and has been studied extensively. Mainstream algorithms, such as Gated Recurrent Unit (GRU) models, primarily focus on encoding historical factual features within TKGs, often neglecting the importance of incorporating entities and relational features during decoding. This bias ultimately leads to loss of detail and inadequate prediction accuracy during the inference process. To address this issue, a novel ChronoBridge framework is proposed that features a dual mechanism of a chronological node encoder and a bridged feature fusion decoder. Specifically, the chronological node encoder employs an advanced recursive neural network with an enhanced GRU in an autoregressive manner to model historical KG sequences, thereby accurately capturing entity changes over time and significantly enhancing the model’s ability to identify and encode temporal patterns of facts across the timeline. Meanwhile, the bridged feature fusion decoder utilizes a new variant of GRU and a multilayer perception mechanism during the prediction phase to extract entity and relation features and fuse them for inference, thereby strengthening the reasoning capabilities of the model for future events. Testing on three standard datasets showed significant improvements, with a 25.21% increase in MRR accuracy and a 39.38% enhancement in relation inference. This advancement not only improves the understanding of temporal evolution in knowledge graphs but also sets a foundation for future research and applications of TKG reasoning.