Continuous Parameter Research Articles

TiO2/SO4/Ni@SBA-15 catalysts for the selective oxidation of veratryl alcohol to veratraldehyde in a continuous reactor

Veratraldehyde is a crucial industrial component used to make important chemicals, drugs, flavours, perfumes, mosquito repellents and cancer-fighting substances. This research focuses on the continuous processing of veratryl alcohol to veratraldehyde using the catalyst system composed of SBA-15/TiO2/SO4/Ni. The objective of this study is to explore variations in continuous processing parameters and catalyst loading to achieve maximum conversion and selectivity of veratraldehyde. The catalyst was synthesized and then characterized using the X-ray diffraction (XRD), Fourier Transform infrared spectroscopy (FT-IR), Brunauer-Emmett-Teller (BET) surface area analysis, and transmission electron microscopy (TEM), and Temperature Programmed Desorption (TPD). The catalyst was successfully synthesized, and the results of these characterization procedures revealed its particular porosity structure and composition. By attaining maximum conversion and selectivity at a lower temperature of 90 °C, the continuous process highlighted the catalyst's energy efficiency. The findings demonstrated that the highest veratraldehyde selectivity and yield were obtained under optimal reaction conditions, including a reaction time of 2 h for the catalyst SBA-15/TiO2/SO4/Ni(8%). This catalyst produced the best results, with a veratraldehyde selectivity and conversion of 100 and 99.5%, respectively. In addition to optimising the reaction conditions, the catalyst's reusability was assessed by running the same catalyst through the reaction seven times. The outcomes demonstrated that even after numerous applications, the catalyst maintained good activity and selectivity. This shows the promise of the SBA-15/ TiO2/SO4/Ni(8) catalyst system for the selective oxidation of veratryl alcohol to veratraldehyde in terms of excellent selectivity, good yield, and reusability. Veratraldehyde could be produced industrially at low cost and with minimal impact on the environment thanks to this investigation of a continuous process.

Continuous glucose monitoring parameters in pregnancy-related complications in patients with type 1 diabetes: a retrospective cohort study.

Continuous glucose monitoring (CGM) improves pregnancy outcomes in patients with type 1 diabetes (T1D). The primary study objective was to analyze associations between numerous novel CGM parameters and neonatal complications, such as large‑for‑gestational‑age (LGA) neonates, hypoglycemia, hyperbilirubinemia, transient breathing disorders, preterm births, as well as pre‑eclampsia. In this single‑center retrospective cohort study, we recruited 102 eligible pregnant women with T1D who were treated with sensor‑augmented pumps with suspend‑before‑low function from the first trimester. The pregnant patients were admitted for at least 1 control hospital visit in each trimester of gestation for anthropometric and laboratory measurements and collection of sensor data. The median (interquartile range) percentage values for glycated hemoglobin (HbA1c) (first trimester, 6.23 [5.91-6.9]; second trimester, 5.49 [5.16-5.9]; third trimester, 5.75 [5.39-6.29]) and for time‑in‑range (first trimester, 72.4 [67.3-80.3]; second trimester, 72.5 [64.7-79.6]; third trimester, 75.9 [67.1-81.4] met the criteria of well‑controlled T1D in each trimester of pregnancy. Nonetheless, we noted 27% of LGA births, 25% of neonatal hypoglycemia, 33% of hyperbilirubinemia, and 13% of preterm births. Worse glycemic control and more glycemic fluctuations in the second and third trimesters were mainly associated with increased risk of LGA at birth, transient breathing disorders, and hyperbilirubinemia. CGM parameters (mean of daily differences, high blood glucose index, glycemic risk assessment in diabetes equation, or continuous overall net glycemic action) in the patients with T1D are significantly associated with the increased risk of LGA at birth and neonatal transient breathing disorders and hyperbilirubinemia. However, we did not find evidence that novel CGM indices could be more effective in predicting those events than the commonly used CGM parameters or HbA1c levels.

Optimizing trajectories for highway driving with offline reinforcement learning

Achieving feasible, smooth and efficient trajectories for autonomous vehicles which appropriately take into account the long-term future while planning, has been a long-standing challenge. Several approaches have been considered, roughly falling under two categories: rule-based and learning-based approaches. The rule-based approaches, while guaranteeing safety and feasibility, fall short when it comes to long-term planning and generalization. The learning-based approaches are able to account for long-term planning and generalization to unseen situations, but may fail to achieve smoothness, safety and the feasibility which rule-based approaches ensure. Hence, combining the two approaches is an evident step towards yielding the best compromise out of both. We propose a Reinforcement Learning-based approach, which learns target trajectory parameters for fully autonomous driving on highways. The trained agent outputs continuous trajectory parameters based on which a feasible polynomial-based trajectory is generated and executed. We compare the performance of our agent against four other highway driving agents. The experiments are conducted in the Sumo simulator, taking into consideration various realistic, dynamically changing highway scenarios, including surrounding vehicles with different driver behaviors. We demonstrate that our offline trained agent, with randomly collected data, learns to drive smoothly, achieving velocities as close as possible to the desired velocity, while outperforming the other agents.

A randomized method for the identification of switched NARX systems

The identification of switched systems is a complex optimization problem that involves both continuous (parametrizations of the local models, a.k.a. modes) and discrete variables (model structures, switching signal). In particular, the combinatorial complexity associated with the estimation of the switching signal grows exponentially with the number of samples, which makes data segmentation (i.e. estimating the number and location of mode switchings, and the mode sequence) a challenging problem. In this work, we extend a previously developed randomized approach for the identification of switched systems to encompass the estimation of the switching locations. The method operates by extracting samples from a probability distribution of switched models, and gathering information from the associated model performances to update the distribution, until convergence to a limit distribution associated to a specific model. A suitable probability distribution is employed to represent the likelihood of a mode switching at a certain time, and the update process is designed to correct the switching locations and remove redundant switchings. The proposed algorithm has been compared to existing state-of-the-art methods and has been tested on various benchmark examples, to demonstrate its effectiveness.

Multi-dimensional spectral graph wavelet transform

In this paper, we introduce the two-dimensional spectral graph wavelet transform (SGWT) of discrete functions defined on weighted Cartesian product graphs. The graphs we consider are undirected, non-planar, and can be cyclic with no multiple loops and no multiple edges. We build this transform with the help of spectral decomposition of $$N_1N_2 \times N_1N_2$$ Laplacian matrix $$\mathcal {L}$$ of the Cartesian product graph $$G=G_1\square G_2$$ , where $$N_1N_2$$ is the number of vertices of the Cartesian product graph. We have established the inversion formula for SGWT for continuous scale parameters $$s_1$$ and $$s_2$$ . By sampling the scale parameters at discrete values, we obtain discrete SGWT coefficients at different scales and vertices. We have obtained the frame bounds of the frame formed by the set of scaling and wavelet coefficients. Further, we have proved the localization result which holds for both normalized and non-normalized form of Laplacian. Towards the end, we have given the implementation details of SGWT through an example to show how to calculate the graph wavelet coefficients.

Abstract #1406346: Efficacy and Safety of Glucagon-Like Peptide-1 Receptor Agonist (GLP-1-RA) in Patients with Diabetes who Underwent Kidney Transplantation (KT): A Retrospective Study with a Control Group

Glucose like peptide-1 receptor agonist (GLP1-RA) have been shown to reduce cardiovascular mortality and promote weight loss in general population. Uncontrolled diabetes after solid-organ transplantation has been associated to weight gain, high cardiovascular mortality and transplant rejection. Despite the metabolic benefits of GLP1RA, their efficacy and safety in the post-transplant period are yet to be established. Tolerability of GLP1RA during the first months after KT have been questioned due to possible gastrointestinal side effects. We aim to describe the efficacy and safety in controlling diabetes and weight, comparing kidney-transplanted participants with diabetes who initiated GLP1RA to participants who did not. Adults (n=50) with diabetes who underwent KT from 08/2020 to 08/2022 at UAB-Transplant Institute were included. 25 participants on insulin ± oral antidiabetic medications initiated GLP1RA (cases) and 25 participants on insulin ± oral antidiabetic did not start GLP1RA (controls). GLP1RA was started 7.72 months ±5.26 after KT and 88% of them started using GLP1RA for the first time. Participants who started GLP1RA did not report active gastrointestinal symptoms and were educated to eat small-meal portions. Metabolic outcomes evaluated were HbA1C, fructosamine, weight, total daily insulin dose (TDD), and Continuous Glucose Monitoring parameters. The outcomes were evaluated before starting GLP1RA and 6 months after. Due to the longitudinal nature of the study, linear mixed models were developed to test for changes over time. KT participants (case, control) were predominantly of age (56.28 years, 57.40 years (p=0.6)), male (64%, 64% (p=1.0)), African American (60%, 84% (p=0.23)) and had type 2 diabetes (56%, 68% (p=0.1)). There was a statistically significant reduction in weight and total daily insulin dose between the two groups 1) Weight: Cases (at baseline, 206.04±38.09 pounds; at 6 months after starting GLP1RA, 204.17±40.08 pounds). Controls (at baseline, 180.96±45.49 pounds; at 6 months 188.61±43.69 pound) (p=0.001); 2) TDD: Cases (at baseline, 61.28±42.18 units, at 6 months, 56.13 ±40.95 units), Controls (at baseline, 34.03±22.49 units; at 6 months, 41.65±25.58 units) (p=0.001). There was no statistical significance between cases and controls for A1C (p=0.70), fructosamine (p=0.91), Glucose Management Indicator (p=0.91), Time In-Range (p=0.51), Time Below-Rage (p=0.91), blood tacrolimus levels (p=0.91). P-values for the test of interaction between covariate (race, GLP1RA type and insulin type) and time for the outcome were not significant. None of the participants developed significant gastrointestinal side effects and the rate of GLP1RA discontinuation was 0%. GLP1RA are safe and effective in reducing insulin requirements and promoting weight loss in the early KT period. No changes in blood tacrolimus levels were found. There were no differences in glucose control parameters in participants who used GLP1RA compared to other antidiabetic therapies that include insulin. Choosing the ideal patient and offering education prior starting GLP1RA is recommended.

Continuous Mean Distance of a Weighted Graph

We study the concept of the continuous mean distance of a weighted graph. For connected unweighted graphs, the mean distance can be defined as the arithmetic mean of the distances between all pairs of vertices. This parameter provides a natural measure of the compactness of the graph, and has been intensively studied, together with several variants, including its version for weighted graphs. The continuous analog of the (discrete) mean distance is the mean of the distances between all pairs of points on the edges of the graph. Despite being a very natural generalization, to the best of our knowledge this concept has been barely studied, since the jump from discrete to continuous implies having to deal with an infinite number of distances, something that increases the difficulty of the parameter. In this paper, we show that the continuous mean distance of a weighted graph can be computed in time roughly quadratic in the number of edges, by two different methods that apply fundamental concepts in discrete algorithms and computational geometry. We also present structural results that allow for a faster computation of this continuous parameter for several classes of weighted graphs. Finally, we study the relation between the (discrete) mean distance and its continuous counterpart, mainly focusing on the relevant question of convergence when iteratively subdividing the edges of the weighted graph.

Full Bayesian identification of linear dynamic systems using stable kernels

System identification learns mathematical models of dynamic systems starting from input-output data. Despite its long history, such research area is still extremely active. New challenges are posed by identification of complex physical processes given by the interconnection of dynamic systems. Examples arise in biology and industry, e.g., in the study of brain dynamics or sensor networks. In the last years, regularized kernel-based identification, with inspiration from machine learning, has emerged as an interesting alternative to the classical approach commonly adopted in the literature. In the linear setting, it uses the class of stable kernels to include fundamental features of physical dynamical systems, e.g., smooth exponential decay of impulse responses. Such class includes also unknown continuous parameters, called hyperparameters, which play a similar role as the model discrete order in controlling complexity. In this paper, we develop a linear system identification procedure by casting stable kernels in a full Bayesian framework. Our models incorporate hyperparameters uncertainty and consist of a mixture of dynamic systems over a continuum spectrum of dimensions. They are obtained by overcoming drawbacks related to classical Markov chain Monte Carlo schemes that, when applied to stable kernels, are proved to become nearly reducible (i.e., unable to reconstruct posteriors of interest in reasonable time). Numerical experiments show that full Bayes frequently outperforms the state-of-the-art results on typical benchmark problems. Two real applications related to brain dynamics (neural activity) and sensor networks are also included.

Targeted Discovery of Low-Coordinated Crystal Structures via Tunable Particle Interactions

Particles interacting via isotropic, multiwell pair potentials have been shown to self-assemble into a range of crystal structures, yet how the characteristics of the underlying interaction potential give rise to the resultant structure remains largely unknown. We have thus developed a functional form for the interaction potential in which all features can be tuned independently. We perform continuous parameter space searches by systematically changing pairs of parameters, controlling the various features of the interaction potential. By enforcing a repulsive first well (controlling particle interactions of the first neighbor shell), we stimulate the formation of low-coordinated assemblies. We report the self-assembly of 20 previously unknown crystal structure types, 14 of which have low coordination numbers. Despite limiting the search to a small region of the vast parameter space of possible particle interactions, a wealth of complexity and symmetry is apparent within these crystal structures, which include clathrates with empty cages and low-symmetry structures. Our findings suggest that an unknown number of previously undiscovered crystal structure configurations are possible through self-assembly, which can serve as interesting design targets for soft condensed matter synthesis.

A first order continuous timeVARwith random coefficients

This article considers a first order continuous time vector autoregression with random coefficients. We discuss some difficulties that arise when the exact discrete analogue is used for estimating the continuous time parameters and provide an estimation method based on an approximate discrete model. Some expressions for the estimator of the drift parameter matrix, for its approximated bias and for the covariance matrix of the parameter estimates are derived. The finite sample performance of the proposed method is studied by a Monte Carlo experiment. We also illustrate the advantages of our model in an application on the expectations theory of the term structure of interest rates. Results show that the performance of the proposed methodology is good, and allowing for time variation on coefficients results in large reductions in the root mean square error of the parameter estimates.

A Randomized Clinical Trial of Transgender Women Switching to B/F/TAF: The (mo)BETTA Trial.

Cardiometabolic disease in transgender women (TW) is affected by gender-affirming hormonal therapies (GAHTs), HIV, and antiretroviral therapy (ART). We evaluated the 48-week safety/tolerability of switching to bictegravir/emtricitabine/tenofovir alafenamide (B/F/TAF) vs continued ART in TW on GAHT. TW on GAHT and suppressive ART were randomized 1:1 to switch to B/F/TAF (Arm A) or continue current ART (Arm B). Cardiometabolic biomarkers, sex hormones, bone mineral density (BMD) and lean/fat mass by DXA scan, and hepatic fat (controlled continuation parameter [CAP]) were measured. Wilcoxon rank-sum/signed-rank and χ2 tests compared continuous and categorical variables. TW (Arm A n = 12, Arm B n = 9) had a median age of 45 years. Ninety-five percent were non-White; 70% were on elvitegravir or dolutegravir, 57% TAF, 24% abacavir, and 19% TDF; 29% had hypertension, 5% diabetes, and 62% dyslipidemia. There were no adverse events. Arm A/B had 91%/89% undetectable HIV-1 RNA at week 48 (w48). Baseline (BL) osteopenia (Arm A/B 42%/25%) and osteoporosis (17%/13%) were common, without significant changes. BL lean/fat mass were similar. At w48, Arm A had stable lean mass but increased limb (3 lbs) and trunk (3 lbs) fat (within-arm P < .05); fat in Arm B remained stable. No changes occurred in lipid or glucose profiles. Arm B had a greater w48 decrease (-25 vs -3 dB/m; P = .03) in CAP. BL and w48 concentrations of all biomarkers were similar. In this cohort of TW, switch to B/F/TAF was safe and metabolically neutral, though greater fat gain occurred on B/F/TAF. Further study is needed to better understand cardiometabolic disease burden in TW with HIV.

Comparison of hemodynamic consequences of hand ventilation versus machine ventilation for transportation of post-operative pediatric cardiac patients.

Hemodynamic monitoring during in-hospital transport of intubated patients is vital; however, no prospective randomized trials have evaluated the hemodynamic consequences of hand versus machine ventilation during transport among pediatric patients' post-cardiac surgery. The authors hypothesized that manual ventilation after pediatric cardiac surgery would alter hemodynamic and arterial blood gas (ABG) parameters during transport compared to mechanical ventilation. A prospective randomized trial. Tertiary cardiac care hospital. Pediatric cardiac surgery patients. One hundred intubated pediatric patients were randomized to hand or machine ventilation immediately post-cardiac surgery during transport from the operating room to the pediatric post-operative intensive care unit (PICU). Hemodynamic variables, including end-tidal CO2 (ETCO2), oxygen saturation, heart rate, systolic blood pressure (SBP), diastolic blood pressure (DBP), peak airway pressure (Ppeak), and mean airway pressure (Pmean), were measured at origin, during transport, and at the destination. ABG was measured before and upon arrival in the PICU, and adverse events were recorded. The Chi-square test and independent t-test were used for comparison of categorical and continuous parameters, respectively. The mean transport time was comparable between hand-ventilated (5.77 ± 1.46 min) and machine-ventilated (5.96 ± 1.19 min) groups (P = 0.47). ETCO2 consistently dropped during transport and after shifting in the hand-ventilated group, with significantly higher ETCO2 excursion than in machine-ventilated patients (P < 0.05). SBP and DBP significantly decreased during transport (at 5 and 6 min intervals) and after shifting in hand-ventilated patients than in the other group (P < 0.05). Additionally, after shifting, a significant increase in Ppeak (P < 0.001), Pmean (P < 0.001), and pH (P < 0.001), and a decrease in pCO2 (P = 0.0072) was observed in hand-ventilated patients than machine-ventilated patients. No adverse event was noted during either mode of ventilation. Hand ventilation leads to more significant variation in ABG and hemodynamic parameters than machine ventilation in pediatric patients during transport post-cardiac surgery. Therefore, using a mechanical ventilator is the preferred method for transporting post-operative pediatric cardiac patients.

A Scoping Review of Cerebral Doppler Arterial Waveforms in Infants.

Cerebral Doppler ultrasound has been an important tool in pediatric diagnostics and prognostics for decades. Although the Doppler spectrum can provide detailed information on cerebral perfusion, the measured spectrum is often reduced to simple numerical parameters. To help pediatric clinicians recognize the visual characteristics of disease-associated Doppler spectra and identify possible areas for future research, a scoping review of primary studies on cerebral Doppler arterial waveforms in infants was performed. A systematic search in three online bibliographic databases yielded 4898 unique records. Among these, 179 studies included cerebral Doppler spectra for at least five infants below 1 y of age. The studies describe variations in the cerebral waveforms related to physiological changes (43%), pathology (62%) and medical interventions (40%). Characteristics were typically reported as resistance index (64%), peak systolic velocity (43%) or end-diastolic velocity (39%). Most studies focused on the anterior (59%) and middle (42%) cerebral arteries. Our review highlights the need for a more standardized terminology to describe cerebral velocity waveforms and for precise definitions of Doppler parameters. We provide a list of reporting variables that may facilitate unambiguous reports. Future studies may gain from combining multiple Doppler parameters to use more of the information encoded in the Doppler spectrum, investigating the full spectrum itself and using the possibilities for long-term monitoring with Doppler ultrasound.

Machine Learning Assisted Analysis of Electrochemical H2O2 Production

The two-electron water oxidation reaction (2e– WOR) provides an option for on-site production of the valuable chemical hydrogen peroxide (H2O2). A major challenge for the 2e– WOR is the need to improve its selectivity toward H2O2 over O2 production through the oxygen evolution reaction. Electrolyte engineering has shown great promise in improving the selectivity and production rates toward H2O2. However, experimental efforts in optimizing the electrolyte only yield results at discrete conditions, which does not provide a complete picture of performance over the entire parameter range. Here, we apply a machine learning assisted prediction to map out the performance of electrochemical H2O2 production over the continuous parameter space of electrolyte composition and applied potential. We collected experimental data of faradaic efficiencies (FEH2O2) and current densities toward H2O2 (JH2O2) as functions of the applied potential and carbonate ion mole fractions (HCO3– and CO32–) in the electrolyte. The data were then used to train a support vector regression model with 5-fold cross-validation. The accuracy of the model was verified against additional experimental results. The model identified that a maximum H2O2 current density of 2.16 mA/cm2 can be achieved from an optimized bicarbonate ion mole fraction of 0.225 at an applied potential of 3.25 V vs RHE. Lastly, the continuous model enables us to evaluate the thermal efficiency, H2O2 electricity cost, and time needed to produce a set amount of H2O2 over a broad range of electrolytes and applied potential conditions. This work demonstrates how machine learning enables the use of discrete experimental points to construct a continuous picture of electrochemical H2O2 production with high accuracy.

Qualitative and Quantitative Evaluation of Multivariate Time-Series Synthetic Data Generated Using MTS-TGAN: A Novel Approach

To obtain high performance, generalization, and accuracy in machine learning applications, such as prediction or anomaly detection, large datasets are a necessary prerequisite. Moreover, the collection of data is time-consuming, difficult, and expensive for many imbalanced or small datasets. These challenges are evident in collecting data for financial and banking services, pharmaceuticals and healthcare, manufacturing and the automobile, robotics car, sensor time-series data, and many more. To overcome the challenges of data collection, researchers in many domains are becoming more and more interested in the development or generation of synthetic data. Generating synthetic time-series data is far more complicated and expensive than generating synthetic tabular data. The primary objective of the paper is to generate multivariate time-series data (for continuous and mixed parameters) that are comparable and evaluated with real multivariate time-series synthetic data. After being trained to produce such data, a novel GAN architecture named as MTS-TGAN is proposed and then assessed using both qualitative measures namely t-SNE, PCA, discriminative and predictive scores as well as quantitative measures, for which an RNN model is implemented, which calculates MAE and MSLE scores for three training phases; Train Real Test Real, Train Real Test Synthetic and Train Synthetic Test Real. The model is able to reduce the overall error up to 13% and 10% in predictive and discriminative scores, respectively. The research’s objectives are met, and the outcomes demonstrate that MTS-TGAN is able to pick up on the distribution and underlying knowledge included in the attributes of the real data and it can serve as a starting point for additional research in the respective area.

Design and Analysis of Multifidelity Finite Element Simulations

Abstract The numerical accuracy of finite element analysis (FEA) depends on the number of finite elements used in the discretization of the space, which can be varied using the mesh size. The larger the number of elements, the more accurate the results are. However, the computational cost increases with the number of elements. In current practice, the experimenter chooses a mesh size that is expected to produce a reasonably accurate result, and for which the computer simulation can be completed in a reasonable amount of time. Improvements to this approach have been proposed using multifidelity modeling by choosing two or three mesh sizes. However, mesh size is a continuous parameter, and therefore, multifidelity simulations can be performed easily by choosing a different value for the mesh size for each of the simulations. In this article, we develop a method to optimally find the mesh sizes for each simulation and satisfy the same time constraints as a single or a double mesh size experiment. A range of different mesh sizes used in the proposed method allows one to fit multifidelity models more reliably and predict the outcome when meshes approach infinitesimally small, which is impossible to achieve in actual simulations. We illustrate our approach using an analytical function and a cantilever beam finite element analysis experiment.

Two-Phase Flow Coordination Characteristics of H62 Brass Alloy Prepared by Up-Drawing Continuous Casting

In this study, the two-phase flow coordination characteristics between α and β phases of H62 brass made by up-drawing continuous casting are investigated based on the upsetting process. An in situ and new research method for two-phase flow is put forward, and the two-phase flow and grain refinement characteristics are observed under different deformation conditions. The results show that α phase flows fast under 400 °C, β phase is pulled and overridden by α phase under this temperature. When the temperature increases to 500 °C, which is higher than β phase transition temperature, the flow velocity of β phase increases, and the deformation of β phase is found to bulge. The flow of β phase is more sensitive to low deformation rates than α phase. The deformation amount has a more significant impact on β phase than α phase, and the deformation of β phase promotes the grain fragmentation and refinement of α phase accompanied by huge β phase bulging obviously. Under the conditions of high temperature, low deformation rate, and large deformation amount, both phase α and β of up-drawing continuous casting brass alloy are broken and the grains are refined. Based on the two-phase flow characteristic, numerical simulation is used to obtain the optimal continuous extrusion parameters of the H-shaped wire of up-drawing continuous casting H62 brass. Then, the optimized complex cross-section wire is prepared by continuous extrusion experiment. This research aims to provide guidance for the complex processing of two-phase alloys.

Fast prediction of indoor airflow distribution inspired by synthetic image generation artificial intelligence.

Prediction of indoor airflow distribution often relies on high-fidelity, computationally intensive computational fluid dynamics (CFD) simulations. Artificial intelligence (AI) models trained by CFD data can be used for fast and accurate prediction of indoor airflow, but current methods have limitations, such as only predicting limited outputs rather than the entire flow field. Furthermore, conventional AI models are not always designed to predict different outputs based on a continuous input range, and instead make predictions for one or a few discrete inputs. This work addresses these gaps using a conditional generative adversarial network (CGAN) model approach, which is inspired by current state-of-the-art AI for synthetic image generation. We create a new Boundary Condition CGAN (BC-CGAN) model by extending the original CGAN model to generate 2D airflow distribution images based on a continuous input parameter, such as a boundary condition. Additionally, we design a novel feature-driven algorithm to strategically generate training data, with the goal of minimizing the amount of computationally expensive data while ensuring training quality of the AI model. The BC-CGAN model is evaluated for two benchmark airflow cases: an isothermal lid-driven cavity flow and a non-isothermal mixed convection flow with a heated box. We also investigate the performance of the BC-CGAN models when training is stopped based on different levels of validation error criteria. The results show that the trained BC-CGAN model can predict the 2D distribution of velocity and temperature with less than 5% relative error and up to about 75,000 times faster when compared to reference CFD simulations. The proposed feature-driven algorithm shows potential for reducing the amount of data and epochs required to train the AI models while maintaining prediction accuracy, particularly when the flow changes non-linearly with respect to an input.

Effects of different 3D QED vertex Ansätze on the critical coupling

We study the semimetal-insulator phase transition in graphene using a Schwinger-Dyson approach. We consider various forms of vertex Ans\"atze to truncate the hierarchy of Schwinger-Dyson equations. We define a Ball-Chiu-type vertex that truncates the equations without violating gauge invariance. We show that there is a family of these vertices, parametrized by a continuous parameter that we call $a$, all of which satisfy the Ward identity. We have calculated the critical coupling of the phase transition using different values of $a$. We have also tested a common approximation in which only the first term in the Ball-Chiu Ansatz is included. This vertex is independent of $a$, and, although it is not gauge invariant, it has been used many times in the literature because of the numerical simplifications it provides. We have found that, with a one-loop photon polarization tensor, the results obtained for the critical coupling from the truncated vertex and the full vertex with $a=1$ agree very well, but other values of $a$ give significantly different results. We have also done a fully self-consistent calculation, in which the photons are backcoupled to the fermion degrees of freedom, for one choice $a=1$. Our results show that when photon dynamics are correctly taken into account, it is no longer true that the truncated vertex and the full Ball-Chiu vertex with $a=1$ agree well. The conclusion is that traditional vertex truncations do not really make sense in a system that does not respect Lorentz invariance, like graphene, and the need to include vertex contributions self-consistently is likely inescapable.

Quantifying the stochasticity of policy parameters in reinforcement learning problems.

The stochastic dynamics of reinforcement learning is studied using a master equationformalism. We consider two different problems-Q learning for a two-agent game and the multiarmed bandit problem with policy gradient as the learning method. The master equationis constructed by introducing a probability distribution over continuous policy parameters or over both continuous policy parameters and discrete state variables (a more advanced case). We use a version of the moment closure approximation to solve for the stochastic dynamics of the models. Our method gives accurate estimates for the mean and the (co)variance of policy variables. For the case of the two-agent game, we find that the variance terms are finite at steady state and derive a system of algebraic equationsfor computing them directly.