Backward Process Research Articles

Low-Dimensional Gradient Helps Out-of-Distribution Detection.

Detecting out-of-distribution (OOD) samples is essential for ensuring the reliability of deep neural networks (DNNs) in real-world scenarios. While previous research has predominantly investigated the disparity between in-distribution (ID) and OOD data through forward information analysis, the discrepancy in parameter gradients during the backward process of DNNs has received insufficient attention. Existing studies on gradient disparities mainly focus on the utilization of gradient norms, neglecting the wealth of information embedded in gradient directions. To bridge this gap, in this paper, we conduct a comprehensive investigation into leveraging the entirety of gradient information for OOD detection. The primary challenge arises from the high dimensionality of gradients due to the large number of network parameters. To solve this problem, we propose performing linear dimension reduction on the gradient using a designated subspace that comprises principal components. This innovative technique enables us to obtain a low-dimensional representation of the gradient with minimal information loss. Subsequently, by integrating the reduced gradient with various existing detection score functions, our approach demonstrates superior performance across a wide range of detection tasks. For instance, on the ImageNet benchmark with ResNet50 model, our method achieves an average reduction of 11.15 % in the false positive rate at 95 % recall (FPR95) compared to the current state-of-the-art approach.

Evaluating the Impact of Retinal Vessel Segmentation Metrics on Retest Reliability in a Clinical Setting: A Comparative Analysis Using AutoMorph.

Current research on artificial intelligence-based fundus photography biomarkers has demonstrated inconsistent results. Consequently, we aimed to evaluate and predict the test-retest reliability of retinal parameters extracted from fundus photography. Two groups of patients were recruited for the study: an intervisit group (n = 28) to assess retest reliability over a period of 1 to 14days and an intravisit group (n = 44) to evaluate retest reliability within a single session. Using AutoMorph, we generated test and retest vessel segmentation maps; measured segmentation map agreement via accuracy, sensitivity, F1 score and Jaccard index; and calculated 76 metrics from each fundus image. The retest reliability of each metric was analyzed in terms of the Spearman correlation coefficient, intraclass correlation coefficient (ICC), and relative percentage change. A linear model with the input variables contrast-to-noise-ratio and fractal dimension, chosen by a P-value-based backward selection process, was developed to predict the median percentage difference on retest per image based on image-quality metrics. This model was trained on the intravisit dataset and validated using the intervisit dataset. In the intervisit group, retest reliability varied between Spearman correlation coefficients of 0.34 and 0.99, ICC values of 0.31 to 0.99, and mean absolute percentage differences of 0.96% to 223.67%. Similarly, in the intravisit group, the retest reliability ranged from Spearman correlation coefficients of 0.55 and 0.96, ICC values of 0.40 to 0.97, and mean percentage differences of 0.49% to 371.23%. Segmentation map accuracy between test and retest never dropped below 97%; the mean F1 scores were 0.85 for the intravisit dataset and 0.82 for the intervisit dataset. The best retest was achieved with disc-width regarding the Spearman correlation coefficient in both datasets. In terms of the Spearman correlation coefficient, the worst retests of the intervisit and intravisit groups were tortuosity density and artery tortuosity density, respectively. The intravisit group exhibited better retest reliability than the intervisit group (P < 0.001). Our linear model, with the two independent variables contrast-to-noise ratio and fractal dimension predicted the median retest reliability per image on its validation dataset, the intervisit group, with an R2 of 0.53 (P < 0.001). Our findings highlight a considerable volatility in the reliability of some retinal biomarkers. Improving retest could allow disease progression modeling in smaller datasets or an individualized treatment approach. Image quality is moderately predictive of retest reliability, and further work is warranted to understand the reasons behind our observations better and thus ensure consistent retest results.

Predictors for length of stay after surgical aortic valve replacement.

Aortic valve replacement improves and prolongs lives of patients with aortic valve disease, but requires significant healthcare resources, which are mainly determined by the length of associated hospital stays. Therefore, this study aims to identify risk factors for extended length of stay after surgical aortic valve replacement. Between 2018 and 2023, 458 consecutive patients underwent isolated surgical aortic valve replacement at our center and were included in our analysis. To identify independent predictors for hospital and intensive care unit stay, multivariable linear regression analysis using backward elimination process was performed. Upon multivariable linear regression, endocarditis [regression coefficient (β) 2.98; 95% confidence interval (CI) 1.51, 4.45; p<0.001)] and prior aortic valve surgery (β 1.72; 95% CI 0.18, 3.26; p=0.029) were associated with prolonged hospital stay. Prior aortic valve surgery was associated with prolonged intensive care unit stay (β 0.99; 95% CI 0.39, 1.59; p=0.001) as well as chronic obstructive pulmonary disease (β 1.61; 95% CI 0.66, 2.55; p=0.001), smaller prosthetic valve sizes (β -0.18; 95% CI -0.30, -0.06; p=0.003), preoperative atrial fibrillation (β 1.06; 95% CI 0.32, 1.79; p=0.005), and reduced left ventricular ejection fraction (β -0.03; 95% CI -0.05, -0.01; p=0.006). Pending further validation, structured programs aiming to accelerate intensive care unit and hospital discharge after surgical aortic valve replacement should focus on patients with prior cardiac surgery, atrial fibrillation, and chronic obstructive pulmonary disease. Surgeons should aim to implant large-diameter valves. Furthermore, the identified predictors should be used to discuss surgical versus transcatheter procedures in the interdisciplinary heart team.

Cultural adaptation and validation of the Sinhala version of the spiritual needs assessment for patients (S-SNAP) questionnaire

BackgroundSpiritual support for patients and caregivers of critically ill patients is associated with improved quality of life. This aspect, however, is not incorporated into the current care pathways in Sri Lanka. The Spiritual Needs Assessment for Patients (SNAP) questionnaire, comprised of 3 domains: psychosocial, spiritual and religious, gives a platform for clinicians to assess the spiritual needs of those patients. This study presents the results of validation of the Sinhala version of the SNAP (S-SNAP) questionnaire.MethodsThe SNAP was translated from English to Sinhala using the standard forward and backward translation process. After verifying the content validity, unambiguity and clarity of items in a focused group discussion, and a pilot study, the pre-final version was tested among 267 volunteers with cancer selected from three state-run cancer care institutions. Data were analysed for internal consistency and item-total correlations. Factor analysis was done using Varimax rotation with Kaiser normalization. A Scree plot was also made to determine the number of factors.ResultsThe mean (SD) age of subjects was 63.2 (11.4) years. The total S-SNAP score ranged from 22 to 88 (maximum 88). The overall Cronbach’s alpha was 0.94 while item-total correlations varied from 0.26 to 0.87. Total SNAP score showed inverse correlations with age, Charleson Comorbidity index and Barthel index while a positive correlation was seen with the Karnofsky performance status scale (p < 0.05). Kaiser-Meyer-Olkein value of 0.92 (P = < 0.001) for Bartlett’s test indicated adequate sampling and non-linearity of factors. The scree plot showed a four-factor structure explaining 76% variation. Meaning of life and relationship with a supernatural being and religious rituals are loaded as 2 different factors. Worries, fears and forgiveness are grouped as the third factor while relaxation, coping and sharing feelings are loaded separately.ConclusionsThe S-SNAP is a reliable and valid tool to assess spiritual suffering among patients with cancers conversant in the Sinhala language.

Machine learning assisted crystallographic reconstruction from atom probe tomographic images

Atom probe tomography (APT) is a powerful technique for three-dimensional (3D) atomic-scale imaging, enabling the accurate analysis on the compositional distribution at the nanoscale. How to accurately reconstruct crystallographic information from APT data, however, is still a great challenge due to the intrinsic nature of the APT technique. In this paper, we propose a novel approach that consists of the modified forward simulation process and the backward machine learning process to recover the tested crystal from APT data. The high-throughput forward simulations on Al single crystals of different orientations generate 10 000 original 3D images and data augmentation is implemented on the original images, resulting in 100 000 3D images. The big data allows the development of deep learning models and three deep learning algorithms of Convolutional Neural Network (CNN), Vision Transformer (ViT), and Variational Autoencoder (VAE) are used in the backward process. After training, the ViT model performs superior than the CNN and VAE models, which can recover the crystalline orientation outstandingly, as evaluated by the coefficient of determinationR2and the Mean Percent Error (MPE), viz.,R2= 0.93 and MPE = 0.43%,R2= 0.97 and MPE = 0.35%, andR2= 0.93 and MPE = 0.77% for the rotation anglesϕ,ψandθ, respectively, on the test dataset. The present work clearly demonstrates the capability of deep learning models in the recovery of the tested crystals from APT data, thereby paving the way for the further development of large artificial intelligent models of APT.

A method for interactive collaboration between different actors: theory of change workshops

Abstract It is well recognised that the involvement of different actors in health research is crucial to the success of projects and interventions, as it increases the likelihood that they will be relevant, impactful and applicable in practice. However, researchers and policy makers often lack tools for the participatory involvement of actors. One way of actively involving actors is the Theory of Change (ToC) method. A ToC is a model of a change process that describes how and why an intervention leads to the intended change or outcome in a given context. It describes the path along which change is expected to occur and provides a framework for investigating whether and how change takes place. It is a systematic methodology that can be used both prospectively and retrospectively for the planning, implementation and evaluation of interventions. The approach recognises that changes in socio-ecological systems are complex and challenging and that the causal processes are often non-linear and intertwined. A ToC workshop begins with the formulation of a clear vision for the desired change and the long-term goal of the intervention. Starting from the long-term goal, the necessary medium-term and short-term goals are then defined in a backward mapping process. Next, indicators, activities and assumptions are defined for each intermediate goal. The results of a ToC workshop can be visualised in a ToC map, which shows the causal pathways, relationships, and connections between the different components of the ToC. This visual representation helps the actors to understand the logic, connections, and assumptions underlying the ToC. This in turn facilitates a shared understanding, collaboration, and ownership among all actors involved. Overall, successful involvement of actors using the ToC method can not only enrich the research process, but ultimately also increase the health benefits of interventions and health policies.

Guidelines for Integrating actionable A-SMART Learning Outcomes into the Backward Design Process.

Learning outcomes are essential in education, guiding both educators and learners towards desired knowledge, skills, and competencies. The backward design process offers a structured approach to curriculum planning, but its integration with actionable, SMART (Specific, Measurable, Achievable, Relevant, Time-bound) learning outcomes needs further exploration. This guide aims to introduce the concept of "A-SMART" learning outcomes and demonstrate their integration into the backward design process, focusing on outcomes that begin with action verbs. The guide outlines a three-stage curriculum planning approach: (i) define desired results, (ii) determine acceptable evidence of learning, and (iii) plan learning activities. It emphasizes the importance of starting with action verbs in formulating learning outcomes, aligning with Stage 1 of backward design and facilitating the transition to Stage 2 (assessment development). By following this guide, educators will acquire tools to develop effective "A-SMART" learning outcomes. This approach immediately advances to Stage 2 of backward design, improving educational practices and ensuring alignment with assessment methods. The guide provides strategies for formulating outcomes that are Action-oriented, Specific, Measurable, Achievable, Relevant, and Time-based. The integration of A-SMART learning outcomes into the backward design process offers a more cohesive and effective educational framework. This approach enhances clarity for learners, provides guidance for instructors, enables more effective assessments, and improves overall learning experiences. The guide also addresses potential challenges in formulating A-SMART outcomes and suggests solutions, including the use of AI tools for inspiration and critical review.

Subgrid-scale model considering the inverse energy cascade using an artificial neural network

For the closure of the subgrid-scale (SGS) stress tensor, an artificial neural network (ANN)-based SGS model that takes account of the inverse energy cascade in isotropic turbulence is developed. The data required for training this ANN-based SGS model are provided by direct numerical simulation of isotropic turbulence with an inverse energy cascade. Two input features, the root mean square of the rate-of-strain tensor and the product of the eigenvalues of the rate-of-strain tensor, are employed to characterize the inverse energy cascade. An a priori test reveals that the ANN-based model adequately predicts the SGS stress tensor in the backward energy transfer process, and the predictive capability of the gradient model is found to be slightly poorer than that of the ANN-based model, while that of the Smagorinsky model is not satisfactory. In comparison with the gradient model, the ANN-based model even predicts a few backward energy transfer events in the stage of excessive energy dissipation. In addition, the off-diagonal component of the SGS stress tensor, rather than the diagonal component, may be intimately associated with the inverse energy cascade. The ANN-based SGS model presented here is expected to provide inspiration for future investigations of the construction of SGS models that take account of the inverse energy cascade.

The Psychometric Properties of the Persian Version of Emotional Abuse Questionnaire

Objective: Intimate partner emotional abuse is a serious issue that can lead to catastrophic outcomes for victims. Emotional abuse involves psychological tactics to control, manipulate, and degrade a person within an intimate relationship. This research aimed to translate the Emotional Abuse Questionnaire (EAQ) developed by Jacobson and Gottman into Persian for use among Iranian university students. Method: The translation of the 66-item EAQ involved a meticulous forward and backward translation process, linguistic matching, and a pilot review. In this cross-sectional study, 346 university students from Rasht, Iran, completed the EAQ. The mean age of participants was 26.78 ± 8.10 years, with most being female (89.0%). Reliability was evaluated using Cronbach's alpha and test-retest analysis while content and face validity were assessed by a panel of experts. Construct validity was examined through confirmatory factor analysis (CFA) and internal consistency measures. Divergent validity was assessed by comparing the EAQ with the ENRICH Marital Satisfaction Scale (EMS). Results: Impact scores for face validity ranged from 2.33 to 3.92, based on respondents' ratings of frequency and importance. Content validity assessment led to removing four items with a CVR below 0.62, resulting in 62 valid items. The EAQ showed strong internal consistency, with a Cronbach’s alpha of 0.97, exceeding the acceptable threshold of 0.70. CFA results confirmed the validity of the second-order factor model of the EAQ (χ2/df = 4.34, CFI = 0.95, RMSEA = 0.098, SRMR = 0.077). The EAQ demonstrated a strong correlation with EMS measures, confirming divergent validity. Conclusion: The Persian version of the EAQ is a reliable instrument for assessing emotional abuse among Iranian university students. Future research should explore the cultural sensitivity of the questionnaire and investigate associations between emotional abuse and other variables of interest, such as mental health outcomes or relationship dynamics. These avenues promise valuable insights into the frequency and effects of emotional abuse across diverse cultural contexts.

Transitioning to the New Normal and Beyond: Maximizing Students’ Active Learning in Task-Supported Curricula

ABSTRACT Amid the lingering impacts of COVID-19, consistently shifting between online and face-to-face educational settings can be particularly challenging for language teachers and learners. Regardless of the learning environment, it is crucial to provide language learners with opportunities to engage in interactive and communicative activities both in and outside the classroom. In the context of a liminal pandemic era, this article aims to share task-supported learning activities developed for Korean learners at the beginner level. The developed task set contains in-class and out-of-class tasks focused on developing the three modes of communication (interpretive, interpersonal, and presentational) and intercultural competence. Given as open educational resources (OER), these collaboratively developed tasks have been piloted across two higher education settings in the United States, demonstrating their sustainability in various learning environments, whether they be online, in person, or hybrid. These tasks that are designed based on the backward design process provide a systematic roadmap for the transition from perception-based tasks to production-based tasks that adhere to the NCSSFL-ACTFL Can-Do Statements (2017). Students’ reflections and instructors’ observations indicate that these tasks and assessment tools are beneficial in developing students’ language capacities, intercultural competencies, and sense of achievement in learning.

Numerical and experimental investigation of extrusion processes from coil

The use of coil material as semi-finished product in sheet-bulk metal forming (SBMF), offers the possibility to use high-speed presses and thus to achieve high output quantities. However, compared to industrially established forming of pre-cut blanks, the reduced geometric accuracy of the parts due to the asymmetrical material flow represents a challenge. Against this background, the aim of this study is to gain an in-depth knowledge on the influence of the process setup on the material flow direction and on the forming of geared functional parts. A combined numerical-experimental approach of each a lateral and a backward extrusion process of a DC04 sheet metal ( t0 = 2 mm) is applied for a fundamental process analysis. For this purpose, process-dependent causes for the asymmetric part forming are analyzed by means of material flow simulations. Subsequently, the numerical results are verified experimentally in terms of the part geometry, the process force and grain structural changes. Based on the numerical and experimental results, cause-effect relations between material flow, part accuracy and tool load are derived. In addition, the transferability of the causes of the asymmetric material flow to processes with different primary material flow directions and part geometries is verified. The findings facilitate the future counteraction of asymmetrical material flow in SBMF from coil, which in turn leads to enhanced component quality.

Questionnaire Translation and Validation of Problem-Focused Coping in Increasing Work to Family Enrichment in Context of Laos

This study aims to translate and validate a questionnaire of problem-focused coping in increasing work to family enrichment in the context of Laos. The questionnaire is translated by using forward and backward translation process with two Lao bilingual translations: then this questionnaire is evaluated through three validity criteria (content validity, face validity, and construct validity) and reliability process. The findings show that the questionnaire translation is available locally, culturally acceptable, and conceptual equivalence for Lao language. Lao questionnaire version is fulfilled acceptable standards of validity and reliability analysis. This study suggests that the Lao questionnaire version is locally accepted to proceed for further actual study in Lao context.

Effect on the physical field of the die during the backward extrusion process of magnesium alloy wheel

The eccentric placement of billets reduces the service life of the dies, and at the same time causes the scrapping of the wheel blanks. However, for axisymmetric forgings, eccentric placement is unavoidable. In this study, the mutual effects between the billet and the die during backward extrusion of magnesium (Mg) alloy wheels are analyzed by physical and numerical simulations, focusing on the effects on the physical field of the die. There were dangerous points in the design of the combined die structure and the lower mating surface of the female die and the bottom die were prone to stress concentration during the filling process of the Mg wheel, which led to the cracking of the die. The eccentric extrusion of the wheel resulted in a slope in the filling height of the rim, and it had a great effect on the physical field and microstructure distribution at the same time, which in turn affected the physical field distribution of the wheel die, resulting in uneven load distribution at the lower rim of the die and reducing the life of the combined die. The combined die structure optimization solution can process the female die into a curved surface along the fracture surface and at the same time process a hollow circular table to cooperate with the bottom die and the female die.

Influence of material in sheet-bulk metal forming from coil

AbstractSheet-bulk metal forming (SBMF) represents an innovative approach for the efficient manufacturing of functional integrated parts from sheet metal by applying bulk forming processes to sheet metal. It combines the advantages of part weight optimization and process chain shortening. In view of the increasing demand for functionally integrated components, the forming from coil offers an economical possibility to produce high output rates. Nevertheless, an underfilling of functional elements as well as an asymmetric part forming are current challenges. In order to apply SBMF from coil in industrial contexts, it is necessary to understand the causes of these process limits and to develop measures for improving the forming of functional elements and thereby push existing forming limits. In this paper, both a lateral and a backward extrusion process from coil for the forming of internal and external geared SBMF parts are investigated to develop a fundamental process understanding. In a combined numerical-experimental approach, the influence of the flow behavior of different materials as well as the impact of the main material flow direction of the process on the dimensional part accuracy is analyzed. This enables the evaluation of fundamental material-related dependencies and influences regarding the geometrical part accuracy, the mechanical component properties and the tool load for forming of the conventional deep-drawing steel DC04 (t0 = 2 mm), a high-strength steel material (HC260LA) and an aluminum alloy (AA6082 T6). Based on these findings, measures for material flow control are identified and investigated to counteract the challenges of SBMF from coil. These findings permit the identification of potential transferability of the identified cause-effect mechanisms to different processes, component geometries and materials. Thereby, in particular, an increase in coil width as well as lower a strain hardening behavior of the material offers the possibility to improve the geometrical accuracy of the components.

An innovative joint-space multi-axis dynamics modeling theory for planetary rover

To address high degree of freedom planetary rover modeling and enhance dynamic performance, an innovative joint-space multi-axis dynamics modeling theory is proposed. The approach aims to streamline the dynamic modeling process by introducing concepts such as topology, natural coordinate system and kinematic iterations. Through an analysis of closed subtrees’ properties and characteristics, an improved Lagrange equation with topology-operable properties is derived and presented. Subsequently, by analyzing the backward iteration process of external forces and torque and eliminating redundant variables, canonical dynamic equations are derived. These equations feature full parameterized modeling and explicit expression. The correctness and accuracy of this method are validated through dynamic simulations of the Mars rover. The proposed method offers the advantage of significantly reducing modeling and calculation complexity, as demonstrated through comparisons with other mature dynamics methods.

Multi-Level Switching Control Scheme for Folding Wing VTOL UAV Based on Dynamic Allocation

A folding wing vertical take-off and landing (VTOL) UAV is capable of transitioning between quadrotor and fixed-wing modes, but significant alterations occur in its dynamics model and maneuvering mode during the transformation process, thereby imposing greater demands on the adaptability of its control system. In this paper, a multi-level switching control scheme based on dynamic allocation is proposed for the deformation stage. Firstly, according to the physical characteristics of the wing folding mechanism, a dynamic model is established. The influence of the incoming flow on the rotors is considered, and the dynamic coupling characteristics in its transition process are analyzed. Secondly, by inverting the changes in rotor position and axial direction, a dynamic allocation algorithm for the rotors is designed. Then, the quadrotor controller and the fixed-wing controller are switched and mixed in multiple loops to form a multi-level switching control scheme. Finally, the simulation results show that the designed multi-level switching control scheme is effective and robust in forward and backward deformation processes, and its anti-interference ability is stronger compared with that of the control scheme without dynamic allocation.

The Spanish Version of the Sussex-Oxford Compassion for Others Scale (SOCS–O) in Nursing Students: Psychometric Properties and Its Relation with Mindfulness

ObjectivesThis study aimed to examine the psychometric properties of the Sussex-Oxford Compassion for Others Scale (SOCS–O) in a large sample of Spanish undergraduate nursing students.MethodAfter a forward–backward translation process, we conducted a cross-sectional study among nursing students in their first year of training at two Spanish universities. The mean age of the participants (n = 683) was 22.74 years old, and 83.46% were women. Together with compassion for others, mindfulness was also assessed.ResultsDescriptive results revealed high scores across all dimensions of the SOCS–O. The data supported a 3-factor structure with correlated factors (χ2(167) = 363.47, p < 0.01; CFI = 0.96; TLI = 0.95; RMSEA = 0.05, 90% CI = 0.04, 0.05; SRMR = 0.04). Reliability estimates, measured by Cronbach’s alpha, were excellent, ranging from 0.87 to 0.93. Measurement invariance across gender was confirmed. Men exhibited significantly lower levels in all compassion dimensions: Recognizing suffering (mean difference = − 0.40), Understanding the universality of suffering (mean difference = − 0.33), and Behavioral compassion (mean difference = − 0.57) (all p < 0.01). The structural equation model showed a clear relationship between mindfulness and compassion (χ2(581) = 1345.79, p < 0.01; CFI = 0.92; TLI = 0.92; RMSEA = 0.04, 90% CI = 0.04, 0.04; SRMR = 0.06).ConclusionsThe findings showed that the Spanish version of the SOCS–O has robust psychometric properties. The SOCS–O is a reliable tool for measuring compassion with three subscales and helps advance understanding of compassion among student nurses. The positive relationship between mindfulness and compassion suggests that incorporating mindfulness training into nursing curricula could enhance the delivery of compassionate care.PreregistrationThis study was not preregistered.

Experimental Study on Oil-Gas Interaction Mechanism during Offshore Heavy Oil Gas Injection.

Offshore heavy oil injection gas extraction is a highly scrutinized area in today's petroleum industry. However, the interaction mechanisms between oil and gas are not clear. To elucidate these mechanisms, an indoor experimental setup was established for research purposes. The effects of different types of gases on heavy oil expansion, mass transfer mechanisms between gas and heavy oil, the influence of gas injection on heavy oil phase behavior, and the testing of minimum miscibility pressures are investigated in this study. The results indicate that CO2 yields the best reduction in the heavy oil viscosity. Both forward and backward multiple contact mass transfer processes demonstrate nonmiscible multiple contact dynamic displacement mechanisms involving CO2 dissolution and condensation, as well as C1 extraction and coextraction. Nonmiscible multiple contact dynamic displacement of natural gas primarily involves limited dissolution and condensation of light hydrocarbon components and intermediate hydrocarbon components, with an extremely weak extraction effect. The minimum miscibility pressures are in the order of CO2 < natural gas < N2. This research provides important experimental evidence and theoretical guidance for further improving offshore heavy oil injection gas technology and practice.

Physical field regulation of magnesium alloy wheel formed by backward extrusion process with multi-stage variable speed

To reduce the tonnage required during the forming process of magnesium (Mg) alloy wheels by backward extrusion (BE) and refine the grains at the wheel bottom, the multi-stage variable (MSV) speed process is proposed based on the characteristics of Mg alloy wheels formed by BE. A numerical model for the BE process of Mg alloy wheels using the Defrom-3D was constructed, and the effects of constant speed and MSV speed processes on the physical fields and DRXed grain size evolution during the forming of Mg alloy wheels were investigated. Lastly, the simulation results were indirectly verified by industrial trial production. The results show that constant speed extrusion has a small effect on the effective stress, but a large effect on the wheel temperature distribution, where the forming speed of the upper rim plays a decisive role in the final temperature distribution of the wheel. MSV speed process is used to regulate the physical field evolution during wheel forming based on the effect of constant speed on the physical field of the wheel. With the high-speed forming of the rim and low-speed forming of the upper wheel rim (9-9-1 mm/s), the temperature and effective stress distribution of the wheel blank are reasonable while reducing the forming tonnage and refining the grains at the wheel bottom.

DGRM: Diffusion-GAN recommendation model to alleviate the mode collapse problem in sparse environments

Generative adversarial network (GAN) has been widely adopted in recommender systems (RSs) to improve the recommendation accuracy. However, existing GAN-based models often suffer from the mode collapse problem in sparse environments and fail to adequately capture the complexity of user preferences and behaviors, which affects recommendation performance. To address these issues, we introduce a diffusion model (DM) into the GAN framework, proposing an efficient Diffusion-GAN recommendation model (DGRM) to achieve mutual enhancement between the two generative models. This model first utilizes the forward process of DM to generate conditional vectors that guide the training of the GAN generator. Subsequently, the backward process of DM assists the GAN discriminator using Wasserstein distance during adversarial training. The Wasserstein distance is adopted to solve the asymmetry of Kullback-Leibler (KL) divergence as a loss function in traditional GANs. Experiments on multiple datasets demonstrate that the proposed model effectively alleviates mode collapse and surpasses other state-of-the-art (SOTA) methods in various evaluation metrics.