Regularization Method Research Articles

Diverse randomized value functions: A provably pessimistic approach for offline reinforcement learning

Offline Reinforcement Learning (RL) faces challenges such as distributional shift and unreliable value estimation, especially for out-of-distribution (OOD) actions. To address these issues, existing uncertainty-based methods penalize the value function with uncertainty quantification and require numerous ensemble networks, leading to computational challenges and suboptimal outcomes. In this paper, we introduce a novel strategy that employs diverse randomized value functions to estimate the posterior distribution of Q-values. This approach provides robust uncertainty quantification and estimates the lower confidence bounds (LCB) of Q-values. By applying moderate value penalties for OOD actions, our method fosters a provably pessimistic approach. We also emphasize diversity within randomized value functions and enhance efficiency by introducing a diversity regularization method, thereby reducing the requisite number of networks. These modules result in reliable value estimation and efficient policy learning from offline data. Theoretical analysis shows that our method recovers the provably efficient LCB-penalty under linear MDP assumptions. Extensive empirical results demonstrate that our proposed method significantly outperforms baseline methods in terms of performance and parametric efficiency.

Correction of nonmonotone trust region algorithm based on a modified diagonal regularized quasi-Newton method

In this paper, a new appropriate diagonal matrix estimation of the Hessian is introduced by minimizing the Byrd and Nocedal function subject to the weak secant equation. The Hessian estimate is used to correct the framework of a nonmonotone trust region algorithm with the regularized quasi-Newton method. Moreover, to counteract the adverse effect of monotonicity, we introduce a new nonmonotone strategy. The global and superlinear convergence of the suggested algorithm is established under some standard conditions. The numerical experiments on unconstrained optimization test functions show that the new algorithm is efficient and robust.

Noncoercive convex sweeping processes with velocity constraints

In this paper, we investigate noncoercive monotone convex sweeping processes with velocity constraints, a topic not previously investigated. Using some fundamental results on Bochner integration, the Tikhonov regularization method, a solution existence result for coercive convex sweeping processes with velocity constraints and a useful fact on measurable single-valued mappings, we prove the solution existence and properties of the solution set of noncoercive convex sweeping processes with velocity constraints under suitable conditions. These results represent a significant contribution, addressing a specific aspect of an open question posed in our recent work [Adly et al., Convex and nonconvex sweeping processes with velocity constraints: well-posedness and insights. Appl Math Optim. 2023;88(Paper No. 45)]. Additionally, we resolve two other open questions from the same paper concerning the behaviour of the regularized trajectories, relying on the Dominated Convergence Theorem for Bochner integration.

Validating and speeding up x-ray tomographic inversions in tokamak plasmas

X-ray tomography is a precious tool in tokamaks that provides rich information about the core plasma, such as local impurity concentration, electron temperature and density as well as magnetic equilibrium (ME) and magnetohydrodynamic activity. Nevertheless, inferring the local plasma emissivity from a sparse set of line-integrated measurements is an ill-posed problem that requires dedicated regularization and validation methods. Besides, speeding up the inversion algorithm in order to be compatible with real-time control systems is a challenging task with traditional approaches. In this contribution, in a first part we introduce tools aiming at validating and speeding up the x-ray tomographic inversions based on Tikhonov regularization, including ME constraint and parameter optimization, taking the WEST geometry as an example. In a second part, an alternative approach compatible with real-time, based on a set of neural networks is proposed and compared with the Tikhonov approach for an experimental case.

Disentangled representation learning with physics-informed variational autoencoder for structural health monitoring

The maintenance and replacement of aging infrastructure is becoming increasingly costly, as a large number of structures is approaching the end of its design lifetime. Labor intensive manual inspection and assessment on a large scale is often infeasible, while data-driven machine learning techniques can fail to identify relevant failure mechanisms and suffer from poor generalization to previously unseen conditions, particularly when data is sparse. We propose a multi-task learning approach based on a physics-informed variational autoencoder that leverages a dictionary of physics-based models to improve the predictive performance when limited data is available. The latent space of the autoencoder is augmented with a set of physics-based latent variables that are interpretable and allow for prior knowledge on the latent variables to be included in the autoencoder formulation. To prevent the data-driven components of the encoder and decoder from overriding the known physics, a regularization method is used to impose physical constraints on the latent space and the generative model prediction. The feasibility of the proposed approach is evaluated on a case study of quay walls in Amsterdam, demonstrating improved performance over purely data-driven methods.

Fisher–KPP-type models of biological invasion: open source computational tools, key concepts and analysis

This review provides open-access computational tools that support a range of mathematical approaches to analyse three related scalar reaction–diffusion models used to study biological invasion. Starting with the classic Fisher–Kolomogorov (Fisher–KPP) model, we illustrate how computational methods can be used to explore time-dependent partial differential equation (PDE) solutions in parallel with phase plane and regular perturbation techniques to explore invading travelling wave solutions moving with dimensionless speed c ≥ 2 . To overcome the lack of a well-defined sharp front in solutions of the Fisher–KPP model, we also review two alternative modelling approaches. The first is the Porous–Fisher model where the linear diffusion term is replaced with a degenerate nonlinear diffusion term. Using phase plane and regular perturbation methods, we explore the distinction between sharp- and smooth-fronted invading travelling waves that move with dimensionless speed c ≥ 1 / 2 . The second alternative approach is to reformulate the Fisher–KPP model as a moving boundary problem on 0 < x < L ( t ) , leading to the Fisher–Stefan model with sharp-fronted travelling wave solutions arising from a PDE model with a linear diffusion term. Time-dependent PDE solutions and phase plane methods show that travelling wave solutions of the Fisher–Stefan model can describe both biological invasion ( c > 0 ) and biological recession ( c < 0 ) . Open source Julia code to replicate all computational results in this review is available on GitHub; we encourage researchers to use this code directly or to adapt the code as required for more complicated models.

Computer modeling of temperature distribution in the course of sintering of glass microspheres

Composite materials obtained by sintering glass microspheres are considered the promising ones for the underwater technology. The high indicators of the hydrostatic strength combined with the heat and sound insulation properties are due to the microspheres maintaining their shapes as close to the spherical as possible. The computer model of the temperature distribution is developed based on the physical and chemical conceptions of the glass composites structure formation by sintering the glass microspheres of sodium borosilicate composition and processing the experimental results of the temperature measurements. The modeling methodology is based on the principles of decomposition of the task by the researcher and ranking the temperature and time parameters. The ratio of the hydrostatic strength to the density of the samples is taken as a functional criterion and the software application is written in the Python language. The experimental setup that is used for sintering the glass microspheres enables to study the temperature fields and thermal deformation processes in real-time conditions. The coefficients of thermal conductivity of sintering are determined by the regular mode method. The results obtained are aimed at solving one of the most important scientific and practical issues of developing technologies for manufacturing glass composites for strategic purpose.

Trichomoniasis Detection Rate Among Female Patients Consulting Different Clinics in Duhok City, Kurdistan Region, Iraq

Background: Trichomoniasis is a highly prevalent sexually transmitted infection which leads to several public health risks such as urethritis, vaginitis, eventually abortion and sterility and it has been associated with human immunodeficiency virus (HIV).Aim of the study: The current cross-sectional study was conducted to measure the rate of Trichomoniasis among female patients consulting different clinics based on direct examination of vaginal swabs, urine samples and cervical discharges specimens.Materials and method: A total of 276 specimens (106 vaginal, 169 urine and 9 cervical) were collected from females aged 18 -45 years consulting four clinics from November, 2022 to May, 2023. The desired patient information was obtained through a designed questionnaire sheet. All specimens processed and examined by direct microscopy following standard protocols. Results were analyzed statistically by Chi-square test and a p value ≥ 0.05 was considered statistically significant.Results: The overall rate of positivity was 9.78% (27 out 276). The detection rates by examining vaginal swab wet mount, urine wet mount and cervical wet mount methods were 14.15% (15 out of 106), 6.21% (10 out of 161) and 22.22% (2 out of 9) respectively. There were almost no significant differences between the infection and socio-demographic variables and methods used.Conclusion: The direct microscopy of vaginal swab or urine specimens still has a useful ability to discover positive cases. In order to increase positivity rates additional more sensitive diagnostic techniques such as regular polymerase chain reaction methods for suspected cases are recommended.

Asteroid-NeRF: A deep-learning method for 3D surface reconstruction of asteroids

Context. Asteroids preserve important information about the origin and evolution of the Solar System. Three-dimensional (3D) surface models of asteroids are essential for exploration missions and scientific research. Regular methods for 3D surface reconstruction of asteroids, such as stereo-photogrammetry (SPG), usually struggle to reconstruct textureless areas and can only generate sparse surface models. Stereo-photoclinometry (SPC) can reconstruct pixel-wise topographic details but its performance depends on the availability of images obtained under different illumination conditions and suffers from uncertainties related to surface reflectance and albedo. Aims. This paper presents Asteroid-NeRF, a novel deep-learning method for 3D surface reconstruction of asteroids that is based on the state-of-the-art neural radiance field (NeRF) method. Methods. Asteroid-NeRF uses a signed distance field (SDF) to reconstruct a 3D surface model from multi-view posed images of an asteroid. In addition, Asteroid-NeRF incorporates appearance embedding to adapt to different illumination conditions and to maintain the geometric consistency of a reconstructed surface, allowing it to deal with the different solar angles and exposure conditions commonly seen in asteroid images. Moreover, Asteroid-NeRF incorporates multi-view photometric consistency to constrain the SDF, enabling optimised reconstruction. Results. Experimental evaluations using actual images of asteroids Itokawa and Bennu demonstrate the promising performance of Asteroid-NeRF, complementing traditional methods such as SPG and SPC. Furthermore, due to the global consistency and pixel-wise training of Asteroid-NeRF, it produces highly detailed surface reconstructions. Asteroid-NeRF offers a new and effective solution for high-resolution 3D surface reconstruction of asteroids that will aid future exploratory missions and scientific research on asteroids.

Broad‐spectrum and sensitive screening of more than 1000 compounds in equine urine using liquid chromatography/high‐resolution mass spectrometry

RationaleTo uphold the integrity of horseracing and equestrian sports, it is critical for an equine doping control laboratory to develop a comprehensive screening method to cover a wide range of target substances at the required detection levels in equine urine.MethodsThe procedure involved the enzymatic hydrolysis of 3 mL urine samples followed by solid‐phase extraction using HF Bond Elut C18 cartridge. The resulting extracts were then separated on a C18 reversed‐phase column and analyzed using liquid chromatography/high‐resolution mass spectrometry (LC/HRMS) in both electrospray ionization positive and negative modes in two separate injections. The analytical data were obtained in full scan and product ion scan (PIS) modes in an 11 min LC run.ResultsThe method can detect 1011 compounds (in both positive and negative ion modes). Over 95% of the target compounds have limits of detections (LODs) ≤10 ng/mL, and more than 50% of the LODs are ≤0.5 ng/mL. The lowest LOD can reach down to 0.01 ng/mL. The applicability of the method was demonstrated by the successful detection of prohibited substances in overseas and domestic equine urine samples.ConclusionsWe have successfully developed a regular screening method for equine urine samples that can detect more than 1000 compounds at sub‐ppb levels in both positive and negative ion modes with full scan and PIS using LC/HRMS. Furthermore, this method can theoretically be expanded to accommodate an unlimited number of prohibited substances in full‐scan mode.

Sedimentation of a Charged Soft Sphere within a Charged Spherical Cavity.

The sedimentation of a soft particle composed of an uncharged hard sphere core and a charged porous surface layer inside a concentric charged spherical cavity full of a symmetric electrolyte solution is analyzed in a quasi-steady state. By using a regular perturbation method with small fixed charge densities of the soft sphere and cavity wall, a set of linearized electrokinetic equations relevant to the fluid velocity field, electrical potential profile, and ionic electrochemical potential energy distributions are solved. A closed-form formula for the sedimentation velocity of the soft sphere is obtained as a function of the ratios of core-to-particle radii, particle-to-cavity radii, particle radius-to-Debye screening length, and particle radius-to-porous layer permeation length. The existence of the surface charge on the cavity wall increases the settling velocity of the charged soft sphere, principally because of the electroosmotic enhancement of fluid recirculation within the cavity induced by the sedimentation potential gradient. When the porous layer space charge and cavity wall surface charge have the same sign, the particle velocity is generally enhanced by the presence of the cavity. When these fixed charges have opposite signs, the particle velocity will be enhanced/reduced by the presence of the cavity if the wall surface charge density is sufficiently large/small relative to the porous layer space charge density in magnitude. The effect of the wall surface charge on the sedimentation of the soft sphere increases with decreases in the ratios of core-to-particle radii, particle-to-cavity radii, and particle radius-to-porous layer permeation length but is not a monotonic function of the ratio of particle radius-to-Debye length.

Three Regularization Methods for A Homogeneous Time Fractional Diffusion Equation with Space - Dependent Diffusivity Coefficient

We studied an inverse problem for a time fractional diffusion equation with space - dependent diffusivity coefficient in the nonhomogeneous case. This problem is ill – posed in Hadamard’s sense. So, a regularization is essential. Using three regularization methods, we get the estimation of the error between the regularized solution and the exact solution

Reduction of finite sampling noise in quantum neural networks

Quantum neural networks (QNNs) use parameterized quantum circuits with data-dependent inputs and generate outputs through the evaluation of expectation values. Calculating these expectation values necessitates repeated circuit evaluations, thus introducing fundamental finite-sampling noise even on error-free quantum computers. We reduce this noise by introducing the variance regularization, a technique for reducing the variance of the expectation value during the quantum model training. This technique requires no additional circuit evaluations if the QNN is properly constructed. Our empirical findings demonstrate the reduced variance speeds up the training and lowers the output noise as well as decreases the number of necessary evaluations of gradient circuits. This regularization method is benchmarked on the regression of multiple functions and the potential energy surface of water. We show that in our examples, it lowers the variance by an order of magnitude on average and leads to a significantly reduced noise level of the QNN. We finally demonstrate QNN training on a real quantum device and evaluate the impact of error mitigation. Here, the optimization is feasible only due to the reduced number of necessary shots in the gradient evaluation resulting from the reduced variance.

Evaluation of the Training Program in Physical Education for Students without Major of Physical Education Hung Vuong University – Phu Tho

Using regular scientific research methods to evaluate the current status of the quality of the physical education program for students at Hung Vuong University on the following aspects: Current status of factors affecting quality of students' physical education program. The results show that students' results in physical education subjects are generally at an average level. The proportion of students achieving good and excellent results is still small. The rate of students failing is still relatively high.

A semi-supervised segmentation method for microscopic hyperspectral pathological images based on multi-consistency learning.

Pathological images are considered the gold standard for clinical diagnosis and cancer grading. Automatic segmentation of pathological images is a fundamental and crucial step in constructing powerful computer-aided diagnostic systems. Medical microscopic hyperspectral pathological images can provide additional spectral information, further distinguishing different chemical components of biological tissues, offering new insights for accurate segmentation of pathological images. However, hyperspectral pathological images have higher resolution and larger area, and their annotation requires more time and clinical experience. The lack of precise annotations limits the progress of research in pathological image segmentation. In this paper, we propose a novel semi-supervised segmentation method for microscopic hyperspectral pathological images based on multi-consistency learning (MCL-Net), which combines consistency regularization methods with pseudo-labeling techniques. The MCL-Net architecture employs a shared encoder and multiple independent decoders. We introduce a Soft-Hard pseudo-label generation strategy in MCL-Net to generate pseudo-labels that are closer to real labels for pathological images. Furthermore, we propose a multi-consistency learning strategy, treating pseudo-labels generated by the Soft-Hard process as real labels, by promoting consistency between predictions of different decoders, enabling the model to learn more sample features. Extensive experiments in this paper demonstrate the effectiveness of the proposed method, providing new insights for the segmentation of microscopic hyperspectral tissue pathology images.

A regularization strategy to a backward problem for general nonlinear parabolic equations in Sobolev spaces

This article is devoted to solving a backward problem for general nonlinear parabolic equations in Sobolev spaces. The problem is hardly solved by computation since it is severely ill‐posed in Hadamard's sense. Using the Fourier transform and the truncation of high‐frequency components, we construct a regularized solution for the problem from the data given inexactly. As usual for ill‐posed problems, the rate of convergence can be obtained under additional smoothness assumptions on the regularity of the exact solution, the so‐called source conditions. By using a source condition in a generalized form, we prove a convergence estimate in a Sobolev space . Corresponding to different levels of the source condition, the convergence rates are improved gradually. We apply our theory to the backward problem with a convection–diffusion operator and a cubic nonlinearity. Finally, the numerical results are presented to test the validity of the proposed regularization method.

Temperature-Dependent Viscosity Analysis of Powell-Eyring Fluid Model during a Roll-over Web Coating Process.

The roll coating method is of considerable significance in several industries, as it is applied practically in the production of paint, the manufacturing of PVC-coated cloth, and the plastic industry. The current study theoretically and computationally analyses the Powell-Eyring fluids with variable viscosity during the non-isothermal roll-over web phenomenon. Based on the lubrication approximation theory (LAT), the problem was formulated. The system of partial differential equations (PDEs) obtained from the mathematical modeling was further simplified to a set of ordinary differential equations (ODEs) using suitable transformations. A regular perturbation method was implemented to obtain the solution in terms of velocity, pressure gradient, pressure, and flow rate per unit width. This study also captures important engineering characteristics such as coating thickness, Nusselt number, shear stress, roll/sheet separating force, and roll-transmitted power to the fluid. Along with a comparison between the present work and published work, both graphical and tabular representations wer made to study the effects of various factors. It was observed that the velocity profile is the decreasing function of non-Newtonian and Reynold viscosity parameters. In addition, the response surface methodology (RSM) was employed to investigate the sensitivity of the shear stress and the Nusselt number.

대학 입학전형 선택의 영향 요인 분석

Objectives The purpose of this study is to analyze the factors that affect the choice of admission types among college students, including personal background, self-efficacy, autonomy, learning motivation, and major selection motivation. Methods To achieve this goal, a survey was conducted on 1,099 students enrolled at A University in Seoul. Factors influencing admission type preferences were analyzed using chi-square tests, one-way ANOVA, and multinomial logistic regression. Results Differences in admission type preferences emerged based on students' backgrounds, including the region and type of their high school, and the highest educational attainment of their guardians. Among personal characteristics, significant disparities in admission-type preferences were observed based on self-efficacy and major selection motivation. In relation to the origin of the high school, as the region shifted to Seoul and the metropolitan area, and as the educational attainment of male guardians increased, along with stronger intrinsic motives for major selection, the likelihood of opting for the regular comprehensive evaluation process increased. Conversely, in cases where the originating high school was located in urban or smaller cities, and when female guardians possessed higher educational qualifications, along with stronger extrinsic motives for major selection, there was an elevated probability of preferring the regular admission method. Conclusions Given the variations in the backgrounds and attributes of students selecting university admission types, universities should take a keen interest not only in the selection process but also in academic and career support.

Design, modeling and optimization of an adhesively bonded ring joint with U-section for ceramic cylindrical pressure hull of deep-sea underwater vehicles

The brittleness of ceramic increases the risk of damage or collapse when the ceramic cylindrical pressure hull is directly connected with a rigid part, especially under high hydrostatic pressure. To solve the above problem, this paper systematically explores the design, modeling and optimization of an adhesively bonded ring joint with U-section (ABRJ-U) to avoid the direct connection of the ceramic hull of a small hadal glider, Petrel, with the hemispherical metal end caps. The ABRJ-U is comprised of a U-shaped metal ring, a gasket, and epoxy adhesive layer. To optimize the design parameters, a mechanical model of ABRJ-U is established. The stress tests prove that the model has better efficiency in optimization, but lowered accuracy compared with regular finite element method. Considering the difficulty of obtaining a more accurate theoretical model, the optimization efficiency of regular finite element method is selected for further improvement by integrating the response surface methodology and multi-objective genetic algorithm. According to the optimization results, our optimization method considers the stress calibration of both the adhesive and gasket, simultaneously achieving a 20.8% reduction in the mass of ABRJ-U. The proposed methods also provide valuable reference for the design, modeling, and optimization of other bonded joints.

The Piston Slap Force Reconstruction of Diesel Engine Using WOA-VMD and Deconvolution.

In a diesel engine, piston slap commonly occurs concurrently with fuel combustion and serves as the main source of excitation. Although combustion pressure can be measured using sensors, determining the slap force is difficult without conducting tests. In this study, we propose a method to identify the slap force of the piston to solve this difficult problem. The traditional VMD algorithm easily receives noise interference, which affects the value of parameter combination [k, α] and thus affects the extraction accuracy of the algorithm. First, we obtain the transfer function between the incentive and vibration response through percussion tests. Secondly, a variational modal decomposition method based on whale algorithm optimization is used to separate the slap response from the surface acceleration of the block. Finally, we calculated the slap force using the deconvolution method. Deconvolution is a typical inverse problem of mathematics, often prone to ill-conditioning, and the singular value decomposition and regularization method is used to overcome this flaw and improve accuracy. The proposed method provides an important means to evaluate the angular distribution of the slap force, identify the shock positions on the piston liner, and determine the peak value of the waveform which helps us analyze the vibration characteristics of the piston and optimize the structural design of the engine.