Loss Function Research Articles

Prediction of the Infrared Absorbance Intensities and Frequencies of Hydrocarbons: A Message Passing Neural Network Approach.

Accurately and efficiently predicting the infrared (IR) spectra of a molecule can provide insights into the structure-properties relationships of molecular species, which has led to a proliferation of machine learning tools designed for this purpose. However, earlier studies have focused primarily on obtaining normalized IR spectra, which limits their potential for a comprehensive analysis of molecular behavior in the IR range. For instance, to fully understand and predict the optical properties, such as the transparency characteristics, it is necessary to predict the molar absorptivity IR spectra instead. Here, we propose a graph-based communicative message passing neural network algorithm that can predict both the peak positions and absolute intensities corresponding to density functional theory calculated molar absorptivities in the IR domain. By modifying existing spectral loss functions, we show that our method is able to predict with DFT-accuracy level the IR molar absorptivities of a series of hydrocarbons containing up to ten carbon atoms and apply the model to a set of larger molecules. We also compare the predicted spectra with those generated by the direct message passing neural network. The results suggest that both algorithms demonstrate similar predictive capabilities for hydrocarbons, indicating that either model could be effectively used in future research on spectral prediction for such systems.

Digitisation of electrocardiographic images enable artificial intelligence-based mortality prediction

Abstract Background Artificial intelligence-based electrocardiogram (AI-ECG) models have shown excellent performance in a vast number of diagnostic and prognostic tasks. One of such tasks, which is critical for patient risk stratification and clinical intervention, is the prediction of adverse events, including death. Although a plethora of existing AI-ECG models have shown their accuracy in mortality prediction, their development and utilization currently rely on digital ECG signals for training and application. Purpose Many hospitals around the world have a significant number of ECGs stored in image formats or printed as paper ECGs. To address this problem, we developed and tested an ECG image digitisation pipeline to extract digital traces from ECG images and compare the performance of AI-ECG models for mortality prediction, under both natively digital and digitised signal formats. Methods A secondary care dataset of over 1 million ECGs (1,163,401 ECGs; 189,539 patients) was used for this analysis, in which both a natively digital signal and a PDF image was available. The ECG digitisation pipeline comprised of coloured PDF image inputs (2200 x 1700), initially processed to extract the ECG traces and grid features. ECG signals were extracted from the traces after cropping/restoration of the images and scaled appropriately based on the detected grid (2.5-sec strips per lead, 0.5 mV per grid block). Subsequently, both digital and digitised signals were identically pre-processed in order to acquire the ECG data fed into the AI model. In particular, a previously successful convolutional neural network architecture with a discrete-time survival loss function was used for mortality prediction. The model was trained under multiple configurations, using either the digital or the digitised ECGs as training/testing datasets (Figure 1). ECGs were split by patient ID to ensure cross-subject generalization with a training/validation/testing split of 50%/10%/40% ratio, respectively. Finally, the models were evaluated by the c-index. Results A comparison of the digital and digitised ECGs on the respective 2.5-sec strips available in both formats revealed a mean absolute error of 0.014 (CI: 0.07 – 0.021) and a Pearson’s R of 0.98 (CI: 0.96 – 1.00) between the signals, demonstrating an accurate ECG digitisation (Figure 2 shows overlapping natively digital (black) and digitised (red) traces). The performance of the digitisation was also evident in the performance of the AI-ECG mortality prediction task, with the model achieving c-indices of 0.775 (CI: 0.774 – 0.776) and 0.772 (CI: 0.771 – 0.774) when trained and tested on digital and digitised signals, respectively. Conclusion We described the first digitisation-based AI-ECG model for mortality prediction. Our results demonstrate the potential for clinical settings that lack the digital ECG infrastructure, to develop and deploy AI models using ECG image formats.Figure 1Figure 2

A novel risk-stratification model that predicts long-term clinical outcomes in patients with recurrent pericarditis

Abstract Background Recurrent Pericarditis (RP) is condition associated with significant debility and morbidity. Previous studies have identified variables such as elevated C-Reactive protein, corticosteroid usage, and presence of delayed hyper-enhancement (DHE) on cardiac magnetic resonance imaging are associated with recurrence. To date however, there is no established model for risk stratifying outcomes in these patients, therefore we attempted to develop a model that could predict long-term outcomes in patients with RP and stratify them into groups based on risk. Methods A total of 365 consecutive patients with RP from 2012-2019 were retrospectively examined. Our primary outcome was clinical remission (CR) defined as symptom resolution in addition to complete cessation of all anti-inflammatory therapy. We employed 5 survival models: CoxPH, RSF, SSVM, GBSA and XGBoost with Cox loss function to determine the likelihood of achieving CR within 5 years and to stratify patients into specific risk groups. We divided the dataset into training (70%) and test (30%) sets. Model training and optimization of the training set occurred via 5-fold cross validation. Model evaluation and selection was based on C-index. Finally, we employed an explainable artificial intelligence approach, SHapley Additive exPlanations to generate both individual and global explanation of the models decision. Results Amongst the cohort, the mean age was 46 ± 15 years, 205 (56%) were female, and the main etiology was idiopathic in 223 (61%). CR was achieved in 118 (32%) patients. The final model included steroid dependency, total number of recurrences, pericardial DHE, age, etiology, gender, ejection fraction and heart rate as its most important parameters. We observed similar CR rates across three risk score ranges and accordingly stratified patients into low (n=62), intermediate (n=154), and high-risk (n=149) groups. Scores of 3 or lower were categorized as low risk, with CR rates exceeding 90%. Scores of 4 to 7 were associated with CR rates of 50%. Scores of 8 or higher were deemed high-risk, with CR rates notably below 10%. Overall, the model predicted the outcome with C-index of 0.800 on the test set and exhibited a significant ability in stratification of patients into low-, intermediate-, and high-risk groups (log-rank test; P <0.0001). Conclusion We created a novel risk prediction model that can aid in predicting rates clinical remission in patients with recurrent pericarditis as well as aid in stratification into low, medium, and high risk groups. This risk prediction model may be helpful in diagnosis and management of these complicated patients.Figure 1

HADNet: A Novel Lightweight Approach for Abnormal Sound Detection on Highway Based on 1D Convolutional Neural Network and Multi-Head Self-Attention Mechanism

Video surveillance is an effective tool for traffic management and safety, but it may face challenges in extreme weather, low visibility, areas outside the monitoring field of view, or during nighttime conditions. Therefore, abnormal sound detection is used in traffic management and safety as an auxiliary tool to complement video surveillance. In this paper, a novel lightweight method for abnormal sound detection based on 1D CNN and Multi-Head Self-Attention Mechanism on the embedded system is proposed, which is named HADNet. First, 1D CNN is employed for local feature extraction, which minimizes information loss from the audio signal during time-frequency conversion and reduces computational complexity. Second, the proposed block based on Multi-Head Self-Attention Mechanism not only effectively mitigates the issue of disappearing gradients, but also enhances detection accuracy. Finally, the joint loss function is employed to detect abnormal audio. This choice helps address issues related to unbalanced training data and class overlap, thereby improving model performance on imbalanced datasets. The proposed HADNet method was evaluated on the MIVIA Road Events and UrbanSound8K datasets. The results demonstrate that the proposed method for abnormal audio detection on embedded systems achieves high accuracy of 99.6% and an efficient detection time of 0.06 s. This approach proves to be robust and suitable for practical applications in traffic management and safety. By addressing the challenges posed by traditional video surveillance methods, HADNet offers a valuable and complementary solution for enhancing safety measures in diverse traffic conditions.

A 4IR-Driven operational risk model for CO2 storage in deepwater abandoned hydrocarbon reservoirs

This study presents the development of an advanced operational risk model that leverages Fourth Industrial Revolution (4IR) technologies to optimize carbon dioxide (CO2) storage within deepwater abandoned hydrocarbon reservoirs. The model systematically combines Artificial Neural Networks (ANN) with the optimization capabilities of Genetic Algorithms (GA) and the probabilistic analysis strengths of a Bayesian Network (BN) to perform dynamic and comprehensive risk assessments. By applying the model to a dataset covering a 200-year timeframe, it effectively forecasts CO2 storage capacities while simultaneously evaluating associated risks across different operational scenarios. One of the key innovations of this model is the introduction of a novel loss function designed to precisely manage forecast deviations and enhance the efficiency of operational processes. This function is critical in ensuring that the model remains robust and accurate in real-time risk assessments, allowing for more reliable decision-making in CO2 storage operations. In addition, the study conducts an economic evaluation that underscores the crucial role of 45Q tax credits in bolstering the financial sustainability of carbon sequestration projects. The analysis highlights how these credits significantly reduce the economic barriers to adopting carbon utilization, storage, and sequestration (CUSS) technologies, making large-scale implementation more feasible. The model's performance is underscored by its ability to achieve a 49% CO2 retention rate over two centuries, with an impressively low average error margin of 0.249%. These results highlight the model's impressive efficiency and accuracy, while also demonstrating its capacity to markedly improve the predictability of CO2 storage outcomes. The findings suggest that this model could play a pivotal role in advancing global sustainability efforts by optimizing CO2 storage processes, thereby contributing to the reduction of atmospheric CO2 levels and supporting long-term climate goals.

DuINet: A dual-branch network with information exchange and perceptual loss for enhanced image denoising

Image denoising is a fundamental task in image processing and low-level computer vision, often necessitating a delicate balance between noise removal and the preservation of fine details. In recent years, deep learning approaches, particularly those utilizing various neural network architectures, have shown significant promise in addressing this challenge. In this study, we propose DuINet, a novel dual-branch network specifically designed to capture complementary aspects of image information. DuINet integrates an information exchange module that facilitates effective feature sharing between the branches, and it incorporates a perceptual loss function aimed at enhancing the visual quality of the denoised images. Extensive experimental results demonstrate that DuINet surpasses existing dual-branch models and several state-of-the-art convolutional neural network (CNN)-based methods, particularly under conditions of severe noise where preserving fine details and textures is critical. Moreover, DuINet maintains competitive performance in terms of the LPIPS index when compared to deeper or larger networks such as Restormer and MIRNet, underscoring its ability to deliver high visual quality in denoised images.

A Novel Technique in Determining Mud Cake Permeability in SiO2 Nanoparticles and KCl Salt Water Based Drilling Fluid using Deep Learning Algorithm

The permeability of the mud cake formed at the formation-wellbore interface is an important factor in the designing of water-based drilling fluids. This study presents a novel approach to utilizing experimental thixotropic and rheological parameters of polymeric water-based drilling fluids having varying concentrations of SiO2 nanoparticles and KCl salt. A fully connected feed-forward multi-layered neural network, more commonly known as a Multilayer Perceptron (MLP) was developed to predict the mud cake permeability using input parameters such as SiO2 & KCl concentration, differential pressure, temperature, mud cake thickness, API LPLT and HPHT filter loss volume and spurt loss volume. The results suggested that the developed Multilayer Perceptron model effectively determined the mud cake permeability based on the input parameters of the WBDF mentioned above. The model converged on the global minima, minimizing the loss function using the Gradient descent algorithm. A higher Coefficient of Determination (R2) value i.e., 0.8781, and a lesser Root Mean Square Error (RMSE) value i.e., 0.04378 indicates the higher accuracy of the model. Pearson’s Coefficient of Correlation obtained via the heatmap indicates that mud cake permeability is strongly influenced by the differential pressure followed by filter loss volume, spurt loss volume, mud cake thickness, and temperature. Previous similar studies have focused on using machine learning algorithms, this study utilized a robust deep learning algorithm i.e., Multilayer Perceptron (MLP) neural network to simultaneously model the combined effects of SiO2 nanoparticles and KCl salt concentrations on mud cake permeability, offering an unprecedented level of accuracy in predicting key WBDF performance parameters

Siamese neural networks can identify subjects from anonymised ECGs

Abstract Background/Introduction Many research databases contain anonymised electrocardiograms (ECGs) linked to other sensitive information. ECGs hold features unique to individuals, potentially enabling subject identification from anonymised ECGs. Purpose We assessed if artificial intelligence approaches to output ECG pair similarity can re-identify individuals from anonymised ECGs. Additionally, we aimed to explore clinical risk prediction using ECG similarity over time. Methods We used a convolutional Siamese neural network (SNN), with a triplet loss function, to train a deep learning model determining if two ECGs belong to the same individual (Figure 1). The model aims to encode ECG inputs from same subjects closer than ECG inputs from different subjects. An ECG similarity score is output, corresponding to the probability two ECGs belong to the same subject. This continuous metric can be used in a binary manner. Below a threshold, two ECGs are classed as from the same individual. Our dataset comprised 72,455 secondary care subjects from a USA cohort with 4 to 75 ECGs each, totalling 864,283 ECGs. These were split 50:10:40% at the subject level for training, validation and testing. Results In 2,689,124 same-subject pairs and 2,689,124 different-subject pairs from the test set, the model achieves an accuracy of 91.68% against the binary threshold. This improves to 93.61% and 95.97% in outpatient and normal ECG subsets, respectively. We tested the model using a subject’s ECG to identify them from a group where only one ECG also belongs to that subject. 89.4% success rates occur in groups of 100 normal ECGs (if random, success would be 1%). 91.4%, 95.1%, and 96.3% success occurs when supplemented with subject gender, age (decade-bracket), and both, respectively. We calculated a model certainty score for this task. 99% success occurs when >95% certain. This is in 65% of group size 100 trials. In groups of 1000 normal ECGs, the success rate is 72.5% with no additional information, and 75.9%, 84.2% and 87.5% when supplemented as before (if random, success would be 0.1%). >95% certainty occurs in 33% of group size 1000 trials. For a given subject, ECG pair similarity proved clinically informative. The model accounts for slight variations within a subject’s ECGs, even when far apart in time (Figure 2A). However, it incorrectly predicts two ECGs as being from different individuals for substantial ECG morphology changes (Figure 2B), which could be used to identify clinical deterioration. The temporal evolution of ECG pair similarity reflects clinical trajectory, with greater dissimilarity associating with all-cause mortality (Hazard ratio, 1.22 per 1 standard deviation change, p < 0.0001). Conclusion(s) Anonymised ECGs retain information useful for subject re-identification. Whilst this implies possible ethical and data protection issues with large datasets, such approaches offer potential for continuous monitoring and identifying clinical deterioration.Figure 1Figure 2

Artificial intelligence-enhanced electrocardiography predicts 10-year risk of atherosclerotic cardiovascular disease

Abstract Background The ACC/AHA pooled cohort equations (PCE) calculates the 10-year primary risk of atherosclerotic cardiovascular disease (ASCVD), an indication to initiate primary prevention statin therapy in international guidelines. The validity of PCE is limited by its overestimation of risk in high-risk cohorts and an ethnic heterogeneity. Electrocardiography (ECG) has not been incorporated in current ASCVD risk estimation tools. Artificial Intelligence (AI)-based models are capable to capture patterns that are informative for cardiovascular prognosis. Purpose We explored the predictive value of a discrete-time survival model in 10-year ASCVD risk in comparison to PCE. Methods We developed an AI-ECG model to predict the 10-year ASCVD risk in an unselected secondary care population from the USA comprised of 1,163,401 ECGs from 189,539 patients. Data was split into training/validation and hold-out test sets by patient ID. Our AI-ECG model uses deep learning with a residual convolutional neural network with a discrete-time survival loss function to output subject-specific survival curves, accounting for both time to event and censorship (i.e., loss to follow up). We compared the performance of out AI-ECG model using a subset of outpatient data (n=4590) of the BIDMC dataset to PCE and ASCVD risk factors including systolic blood pressure, total cholesterol, HDL cholesterol, hypertension, smoking history, diabetes mellitus, and ethnicity. Performances were evaluated using cox survival analysis. Our AI-ECG model was externally validated in the UKB Biobank (UKB), where we compared the performance of our AI-ECG model to predictions of existing AI-ECG models such as the Stanford Estimator of Electrocardiogram Risk (SEER) model. Results Our AI-ECG successfully predicted future ASCVD events in individuals (C-index 0.721 (0.719-0.723), 5-year AUC 0.761 (0.758-0.763), and differentiated between high and low risk group with an adjusted HR of 2.33 (2.26-2.41). Upon internal validation our AI-ECG model showed a superior performance (C-index 0.679 (0.651-0.708)) than PCE (0.605 (0.577-0.634)) and all risk factors combined (C index 0.642 (0.613-0.672)). When combining AI-ECG predictions with all risk factors, the model showed additional predictive value (0.686 (0.657-0.715)) (Figure 1). Our AI-ECG model showed modest predictive value within the UKB dataset (C-index 0.655 (0.637-0.673)), but significantly outperformed the SEER model (C-index 0.547 (0.527-0.567), p <0.0001 for comparison with AIRE). Conclusion We have shown the potential of AI-ECG models in predicting future ASCVD risks. AI-ECG models demonstrated accurate discrete time survival prediction and good classification performance superior to that of the SEER model, as well as PCE and ASCVD risk factors. . Accurate risk assessment of ASCVD may help prevent adverse events in high risk subjects and save unnecessary pharmacological interventios in low risk subjects.Figure 1

Behavioral Cloning Strategies in Steering Angle Prediction: Applications in Mobile Robotics and Autonomous Driving

Artificial neural networks (ANNs) are artificial intelligence techniques that have made autonomous driving more efficient and accurate; however, autonomous driving faces ongoing challenges in the accuracy of decision making based on the analysis of the vehicle environment. A critical task of ANNs is steering angle prediction, which is essential for safe and effective navigation of mobile robots and autonomous vehicles. In this study, to optimize steering angle prediction, NVIDIA’s architecture was adapted and modified along with the implementation of the Swish activation function to train convolutional neural networks (CNNs) by behavioral cloning. The CNN used human driving data obtained from the UDACITY beta simulator and tests in real scenarios, achieving a significant improvement in the loss function during training, indicating a higher efficiency in replicating human driving behavior. The proposed neural network was validated through implementation on a differential drive mobile robot prototype, by means of a comparative analysis of trajectories in autonomous and manual driving modes. This work not only advances the accuracy of steering angle predictions but also provides valuable information for future research and applications in mobile robotics and autonomous driving. The performance results of the model trained with the proposed CNN show improved accuracy in various operational contexts.

Implementing Real-Time Image Processing for Radish Disease Detection Using Hybrid Attention Mechanisms.

This paper developed a radish disease detection system based on a hybrid attention mechanism, significantly enhancing the precision and real-time performance in identifying disease characteristics. By integrating spatial and channel attentions, this system demonstrated superior performance across numerous metrics, particularly achieving 93% precision and 91% accuracy in detecting radish virus disease, outperforming existing technologies. Additionally, the introduction of the hybrid attention mechanism proved its superiority in ablation experiments, showing higher performance compared to standard self-attention and the convolutional block attention module. The study also introduced a hybrid loss function that combines cross-entropy loss and Dice loss, effectively addressing the issue of class imbalance and further enhancing the detection capability for rare diseases. These experimental results not only validate the effectiveness of the proposed method, but also provide robust technical support for the rapid and accurate detection of radish diseases, demonstrating its vast potential in agricultural applications. Future research will continue to optimize the model structure and computational efficiency to accommodate a broader range of agricultural disease detection needs.

FEI-YOLO: A Lightweight Soybean Pod-Type Detection Model

Identifying and statistically analyzing soybean pod types are crucial for seed evaluation and yield estimation. Traditional visual assessment by breeding personnel is time-consuming, labor-intensive, and prone to subjective bias, especially with large datasets. Automatic assessment methods usually struggle with the highly confusing pod types with two and three seeds, affecting the model’s identification accuracy. To address these issues, we propose to improve the standard YOLOv5s object detection model to enhance the differentiation between pod types and to boost the model’s efficiency in prediction. To reduce the number of parameters and the computational load, we propose to introduce the FasterNet Block module in the FasterNet model into the original C3 module, leading to improvements in both detection accuracy and speed. To strengthen the feature extraction and representation for specific targets, the Efficient Multi-Scale Attention (EMA) module is incorporated into the C3 module of the backbone network, improving the identification of similar pod types. The Inner-IoU is combined with the CIoU as the loss function to further enhance detection accuracy and generalization. Experiments comparing FEI-YOLO with the baseline YOLOv5s show that FEI-YOLO achieves an mAP@0.5 of 98.6% and an mAP@0.5:0.95 of 81.1%, with improvements of 1.5% and 1.4%, respectively. Meanwhile, the number of parameters is reduced by 13.2%, and FLOPs decreased by 10.8%, demonstrating the model’s effectiveness and efficiency, enabling rapid and accurate identification of soybean pod types from images.

Statistical inference and data analysis for inverted Kumaraswamy distribution based on maximum ranked set sampling with unequal samples

A very useful modification to ranked set sampling (RSS) that allows a larger set size without significantly increasing ranking errors is the maximum ranked set sampling with unequal samples (MRSSU) approach. This article covers the parameter estimation of the inverted Kumaraswamy distribution using MRSSU and RSS designs. The maximum likelihood and Bayesian estimation techniques are considered. The regarded Bayesian estimation technique is determined in the case of non-informative and informative priors represented by Jeffreys and gamma priors, respectively. Squared error and minimum expected are the two loss functions that are employed. We presented a simulation study to evaluate the performance of the recommended estimations using root mean squared error and relative bias. The Bayes point estimates were computed using the Metropolis–Hastings algorithm. Additional conclusions have been made based on actual geological data regarding the intervals between Kiama Blowhole’s 64 consecutive eruptions. Based on the same number of measured units, the results of simulation and real data analysis showed that MRSSU estimators performed much better than their RSS counterparts.

Image Denoising Using Deblur Generative Adversarial Network Denoising U-Net

Convolutional neural networks (CNNs) are becoming increasingly popular for image denoising. U-Nets, a type of CNN architecture, have been shown to be effective for this task. However, the impact of shallow layers on deeper layers decreases as the depth of the network increases. To address this issue, the authors propose a new image denoising method called DGANDU-Net. DGANDU-Net combines the DeblurGAN design with a U-Net architecture. This combination allows DGANDU-Net to effectively remove noise from images while preserving fine details. The authors also propose the use of two loss functions, mean square error (MSE) and perceptual loss, to improve the performance of DGANDU-Net. MSE is used to learn and improve the extracted features, while perceptual loss is used to produce the final denoised image. The authors evaluate the performance of DGANDU-Net on a variety of noise levels and find that it outperforms other state-of-the-art denoising algorithms in terms of both visual quality and two evaluation indices, including peak signal-to-noise ratio (PSNR) and Structural Similarity Index Measure (SSIM). Specifically, for extremely noisy environments with a noise standard deviation of 75, DGANDU-Net achieves an average PSNR of 37.39[Formula: see text]dB in the test dataset. The authors conclude that DGANDU-Net is a promising new method for image denoising that has the potential to significantly improve the quality of medical images used for diagnosis and treatment.

Enhanced robustness in machine learning: Application of an adaptive robust loss function

In the field of machine learning, the selection of a loss function plays a pivotal role in determining the training dynamics and generalization capability of models. Traditional loss functions such as mean square error (MSE) and cross-entropy are often not equipped to handle noisy and anomalous data effectively, which can result in models that perform poorly in practical scenarios. To address these shortcomings, this study introduces a novel adaptive robust loss function, enhanced by a tunable parameter , which allows for flexibility in adjusting the robustness of the function according to the nature of the data being processed. Our research demonstrates that this new loss function significantly improves the performance of linear regression and multilayer perceptron models, particularly in environments laden with noisy and anomalous data. By adapting the parameter , the function can cater to varying levels of data irregularities, thus enhancing the models accuracy and reliability across diverse and complex data environments. This adaptive mechanism not only offers a substantial theoretical contribution to the understanding of robust loss functions but also provides a practical tool for machine learning practitioners to develop models that are resilient to data imperfections. The implications of this research are profound, suggesting a shift towards more adaptive and robust approaches in machine learning model development.

UFLD enhancements: New loss function and image masking

In the field of autonomous driving, lane detection plays a crucial role in ensuring safety and accuracy in navigation. This paper presents enhancements to the Ultra Fast Lane Detection (UFLD) model by introducing a new loss function that ensures smoother and more continuous lane lines, and incorporating image masking techniques to simulate real-world occlusions. These modifications improve the models robustness, particularly in scenarios with degraded lane markings. Through extensive experimentation on the TuSimple dataset, we demonstrate that the proposed enhancements lead to more accurate and realistic lane detection, especially in challenging environments.

Optimal approximations for the free boundary problems of the space-time fractional Black-Scholes equations using a combined physics-informed neural network

The combined physics-informed neural network is employed to deal with the free boundary problems of fractional Black-Scholes equations. The solution assumption and the loss function are determined, the transfer learning is borrowed, the combined neural network with data enhancement layer is designed, then the classical Black-Scholes model is numerically solved and the comparative analysis of numerical results under different neural networks is made. For further insight into the long-term memory of fluctuation, the free boundary problems of the space-time Black-Scholes equations under Caputo, Caputo-Fabrizio and Atangana-Baleanu-Caputo fractional derivatives are studied. The corresponding empirical analyses are presented and the optimal exercise boundaries of American put option are simulated. The market analysis shows that introducing fractional calculus tools and neural network algorithms into American put option pricing can yield more realistic prediction results. The work provides a viable method for subsequent researchers to study American option pricing using fractional calculus and neural networks combined with true market data and to deal with the free boundary problems in other research fields.

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.

Personalized Super Resolution with Face Prior

Face Super-Resolution (FSR) is a critical technology in computer vision that aims to reconstruct high-resolution facial images from low-resolution inputs. Despite recent advancements, current FSR methods struggle to accurately reconstruct personalized and detailed features. This paper proposes a novel FSR approach that addresses these challenges through a personalized feature extraction and fusion framework. Our method integrates a U-Net based downsampling mechanism to extract individual- specific features from high-resolution reference images, which are then fused with a pre-trained Generative Adversarial Network (GAN) for enhanced reconstruction. We introduce a comprehensive loss function that combines reconstruction, adversarial, facial component, and identity preservation losses to guide the learning process. Extensive experiments on the augmented FFHQ dataset demonstrate that our approach significantly improves the reconstruction of rich facial features, particularly for older individuals, outperforming existing state-of-the-art methods in both quantitative metrics and qualitative visual assessments.

Multi-scale self-attention feature decoupling transfer network-based cross-domain capacity prediction of lithium-ion batteries

Lithium-ion battery capacity prediction is paramount for improving the safety and reliability of energy storage devices. However, accurately predicting capacity continues to pose a challenge, stemming from the diverse range of operational conditions and the lack of capacity labels under unpredictable scenarios. To tackle this problem, a multi-scale self-attention feature decoupling transfer network is proposed for predicting the capacity of lithium-ion batteries across diverse operational conditions. More specifically, multi-scale multi-head self-attentive encoders are designed to adaptively extract multi-scale domain-invariant features by using a series of loss function constraints. Then, transfer learning integrated with domain-invariant features is applied to migrate the knowledge of the source domain to the target domain for cross-domain capacity prediction. Finally, the performance of our methods is authenticated through utilization of two battery dataset, each including three different charging and discharging scenarios. Experimental results indicate that the proposed method effectively extracts domain-invariant features from charging data characterized by varying distributions, thereby mitigating the effects of distributional differences. Compared with the state-of-the-art methods, the proposed method shows excellent accuracy and bustling ability in two datasets.