Levels Of Fidelity Research Articles

Stochastic prior for non-parametric star-formation histories

ABSTRACT The amount of power contained in the variations in galaxy star-formation histories (SFHs) across a range of time-scales encodes key information about the physical processes which modulate star formation. Modelling the SFHs of galaxies as stochastic processes allows the relative importance of different time-scales to be quantified via the power spectral density (PSD). In this paper, we build upon the PSD framework and develop a physically motivated, ‘stochastic’ prior for non-parametric SFHs in the spectral energy distribution (SED)-modelling code prospector. We test this prior in two different regimes: (1) massive, $z = 0.7$ galaxies with both photometry and spectra, analogous to those observed with the LEGA-C survey, and (2) $z = 8$ galaxies with photometry only, analogous to those observed with NIRCam on JWST. We find that it is able to recover key galaxy parameters (e.g. stellar mass, stellar metallicity) to the same level of fidelity as the commonly used continuity prior. Furthermore, the realistic variability information incorporated by the stochastic SFH model allows it to fit the SFHs of galaxies more accurately and precisely than traditional non-parametric models. In fact, the stochastic prior is $\gtrsim 2\times$ more accurate than the continuity prior in measuring the recent star-formation rates (log SFR$_{100}$ and log SFR$_{10}$) of both the $z = 0.7$ and $z = 8$ mock systems. While the PSD parameters of individual galaxies are difficult to constrain, the stochastic prior implementation presented in this work allows for the development of hierarchical models in the future, i.e. simultaneous SED-modelling of an ensemble of galaxies to measure their underlying PSD.

Implementing the Effective Coach2Move Approach for Community-Dwelling Older Adults with Mobility Limitations in Physical Therapist Practice: A Multi-Methods Process Evaluation.

The aims of this study were to evaluate the adoption of and fidelity to the Coach2Move approach; identify differences between physical therapists with a specialization in geriatrics and physical therapists without a specialization; explore if level of adoption and specialization explain variances in effectiveness; and explore group differences in attitudes and experiences with implementation. A multi-methods process evaluation of Coach2Move implementation through 2 education days, 3 peer-assessment meetings, and an adapted electronic health record was performed alongside a cluster randomized stepped-wedge trial comparing regular care physical therapy with Coach2Move. Participants were 36 physical therapists with a specialization (n = 17) and without a specialization (n = 19) who treated 292 community-dwelling older adults. Level of adoption and fidelity were analyzed by comparing pre-implementation scores with scores 1year later. Coach2Move adoption was measured by e-assessment scores and fidelity through health record indicators. The impact of specialization and adoption on health outcomes was examined using a mixed-model ANOVA. Therapists' attitudes and experiences were collected through a survey based on semi-structured interviews. Mean total indicator scores on the e-assessment (adoption of a Coach2Move mindset) increased from baseline 17% (SD = 5%) to 47% (SD = 9%) at follow-up. Physical therapists with a specialization in geriatrics scored higher than physical therapists without (54% [SD = 6%] vs 41% [SD = 6%]). Mean indicator scores on health records (fidelity) increased from 35% (SD = 12%) at baseline to 47% (SD = 15%) at follow-up. Mean scores of physical therapists with a specialization in geriatrics were higher. Level of adoption and specialization (yes/no) did not explain the variance in effectiveness. Physical therapists identified important facilitators, including tailored education and peer-assessment meetings, while adequate reimbursement for the extra time investment was considered a necessity. Different workflows in practices were perceived as a barrier. Implementation led to increased adoption and fidelity of the Coach2Move intervention by physical therapists but shows room for improvement. Attitudes towards the approach and its implementation were positive. Future implementation efforts on adoption could be improved by focusing on a fair compensation structure by third party payers and insurance companies and optimizing organizational and financial context within practices. This study evaluated the implementation of Coach2Move, a clinical reasoning approach designed to increase physical activity and improve self-efficacy in older adults. Overall, the study demonstrates the potential of Coach2Move to be effectively adopted by physical therapists. However, addressing therapist compensation and adapting to practice workflows are important considerations for successful large-scale implementation.

Usability Evaluation Ecological Validity: Is More Always Better?

The ecological validity associated with usability testing of health information technologies (HITs) can affect test results and the predictability of real-world performance. It is, therefore, necessary to identify conditions with the greatest effect on validity. We conducted a comparative analysis of two usability testing conditions. We tested a HIT designed for anesthesiologists to detect pain signals and compared two fidelity levels of ecological validity. We measured the difference in the number and type of use errors identified between high and low-fidelity experimental conditions. We identified the same error types in both test conditions, although the number of errors varied as a function of the condition. The difference in total error counts was relatively modest and not consistent across levels of severity. Increasing ecological validity does not invariably increase the ability to detect use errors. Our findings suggest that low-fidelity tests are an efficient way to identify and mitigate usability issues affecting ease of use, effectiveness, and safety. We believe early low-fidelity testing is an efficient but underused way to maximize the value of usability testing.

Standard medical ethnobotany of Kohistan, North Pakistan

BackgroundThis study was exclusively focused on the documentation and cross-cultural evaluation of ethnomedicinal knowledge (EMK) within the diverse linguistic groups of Kohistan situated between the Himalayan and Hindukush Mountain ranges in the north Pakistan.MethodsData were gathered during the field survey (May 2022 to July 2023) through group conversations, semi-structured interviews, and on-site observation. Venn diagrams were employed to illustrate the comparative assessment of EMK, and different ethnobotanical indices were utilized to examine the data.ResultsA total of 96 wild medicinal plant species (MPs) belonging to 74 genera and 52 botanical families were documented. The most reported MPs belong to the family Polygonaceae (11 species), followed by Asteraceae (9 species) and Lamiaceae (8 species). The ethnomedicinal uses of Leontopodium himalayanum, Pedicularis oederi, Plocama brevifolia, Polypodium sibiricum, Pteridium esculentum, Sambucus wightiana, Solanum cinereum, Teucrium royleanum, Rhodiola integrifolia, Aconitum chasmanthum were reported for the first time in this region. Among the reported taxa herbaceous species were dominated (72%), followed by trees and shrubs (17% and 10%, respectively). Digestive problems (40 taxa and 114 use reports) and skin disorders (19 taxa and 549 use reports) were the most cited disease categories, whereas M. communis, M. longifolia, Ajuga integrifolia, Ziziphus jujuba, and Clematis grata exhibited the highest percentage fidelity levels. Out of 109 documented medicinal uses, a mere 12 were shared across all linguistic groups, and Bateri emerges as a notable outlier with the highest number of medicinal uses. In addition, a significant homogeneity was noted in the reported botanical taxa (61 species) among different linguistic groups. However, since the last decade biocultural heritage of Kohistan is facing multifaceted risks that need urgent attention.ConclusionOur findings could be valuable addition to the existing stock of ethnomedicinal knowledge and may provide ethnopharmacological basis to novel drug discovery for preexisting and emerging diseases prioritizing detailed phytochemical profiling and the evaluation of bioactive potential.

A novel vortex-based velocity sampling method for the actuator-line modeling of floating offshore wind turbines in windmill state

The fluid-dynamic simulation of wind turbine aerodynamics is typically tackled by applying multi-fidelity computational tools. In this context, the so-called actuator line model combines a low-fidelity treatment of the rotor with a high-fidelity resolution of the wake. In this paper, a novel formulation of the actuator line model proposes a vortex-based method to sample the flow around the rotor to rigorously assign the forces imparted by the blades. This new technique is implemented into an in-house code developed within the OpenFOAM environment, and it is validated against wind-tunnel experiments on a laboratory-scale horizontal-axis wind turbine operated in fixed-bottom and floating conditions. The calculations are also compared against multi-fidelity simulations performed, on the same test case, in the frame of the OC6 Phase III project. The simulation results, obtained after a systematic analysis and selection of the model parameters, exhibit a remarkable agreement with the available experiments and place the present code in the proper ranking of fidelity levels, in-between momentum-balance methods and blade-resolved CFD models. Finally, the calculations for surge and pitch platform motions demonstrate the capability of the proposed technique to reliably predict the aerodynamics of turbine rotors in dynamic operation at affordable computational cost.

Unsteady Aerodynamics of Delta Kites for Airborne Wind Energy Under Dynamic Stall Conditions

ABSTRACTThree unsteady aerodynamic tools at different levels of fidelity and computational cost were used to investigate the unsteady aerodynamic behavior of a delta kite applied to airborne wind energy. The first tool is an in‐house unsteady panel method that is fast but delivers low to mid fidelity predictions. The second tool uses the open‐source CFD code SU2 to solve the unsteady Reynolds‐averaged Navier–Stokes equations with the SST turbulence model. At an intermediate level of fidelity, a semiempirical dynamic stall model that combines the panel method with a phenomenological dynamic stall module is proposed. The latter has free parameters that are fine‐tuned with CFD results from the second tool. The research on the dynamic stall model has been inspired by two flight test campaigns suggesting dynamic stall phenomena possibly driven by the periodic variation of the angle of attack (aerodynamic pitching motion) during crosswind maneuvers. The recorded inflow along the flight path was prescribed in the three aerodynamic tools. As expected, the price to pay for the low computational cost of the panel method is its inability to capture the dynamic stall phenomenon. The results from unsteady CFD qualitatively matched the experimental data identifying a leading‐edge vortex that forms and detaches cyclically during the pitching motion. Using RANS data, the semiempirical tool was fined‐tuned to reproduce the dynamic stall behavior, becoming an accurate and fast aerodynamic tool for coupling with any kite flight simulator. Further discussions on the effects of kite aerostructural deflections are included.

Simulation of a scintillator-based fast ion loss detector for steady-state operation in Wendelstein 7-X (invited).

A quantitative theoretical framework has been created to model neutral beam injection and fast ion losses in the Wendelstein 7-X (W7-X) stellarator, including a novel method to develop synthetic diagnostics for fast ion loss detectors (FILDs) of many types, such as scintillating and Faraday Cup FILDs. This is the first time that this has been done in stellarator geometry with this level of fidelity, providing a way for fast ion losses to be predicted more precisely in future stellarator experiments and in W7-X. Simulations of the signal seen by a Faraday Cup FILD have been completed for multiple W7-X plasmas and show close agreement with the measured signals. This method is now applied to an actively water-cooled, scintillator-based FILD, which is currently in development to measure the fast ion loss distribution in W7-X in greater detail. The design makes use of a double slit to measure energy-and-pitch-angle-resolved losses of both co-going and counter-going fast ions. The diagnostic, which can be inserted to different radial positions, has been designed to withstand steady-state heat fluxes of up to 120 kW/m2 along with additional transient heat loads of 100 kW/m2 lasting for up to 20s at a time. Simulations of W7-X standard magnetic configuration show up to 8 × 1013 (s-1 cm-2) ion fluxes onto the sensor from each neutral beam source and no signal from the counter-going slit. These simulations will help inform experimental proposals for future W7-X campaigns after installation of this diagnostic.

Optimization framework for multi-fidelity surrogate model based on adaptive addition strategy—A case study of self-excited oscillation cavity

This study proposes a multi-fidelity efficient global optimization framework for the structural optimization of self-excited oscillation cavity. To construct a high-precision multi-fidelity surrogate model to correlate the structural parameters of a self-excited oscillation cavity with the gas precipitation and energy consumption characteristics by effectively fuzing the information of different fidelity levels, choosing different correlation functions and hyper-parameter estimation methods, and learning the correlation between the data. The optimization framework determines various sampling methods and quantities by calculating the minimum Euclidean distance between sample points and sensitivity index. To enhance computational efficiency, a multi-fidelity sample library is established by utilizing both precise and coarse computational fluid dynamics grids. The expected improvement criterion-based algorithm for global optimization is employed as an additive strategy to incorporate additional data points into the model. This approach considers both local and global search of the model, thereby enhancing sample accuracy while reducing computation time. Moreover, the utilization of the highly generalized Non-dominated Sorting Genetic Algorithm-II (NSGA-II) for identifying the Pareto optimal solution set enhances convergence speed. The proposed optimization framework in this study achieves a remarkable level of model accuracy and provides optimal solutions even with a limited sample size. It can be widely used in engineering optimization problems.

Movement of juvenile migratory birds from settlement to adulthood across the non-breeding range.

Among migratory vertebrates, high levels of fidelity to non-breeding sites during adulthood are common. If occupied sites vary in quality, strong site fidelity can have profound consequences for individual fitness and population demography. Given the prevalence of adult site fidelity, the regions of the non-breeding range to which juveniles first migrate, and the scale of any subsequent movements, are likely to be pivotal in shaping distributions and demographic processes across population ranges. However, inherent difficulties in tracking migratory individuals through early life mean that opportunities to quantify juvenile settlement and movements across non-breeding ranges, and the mechanisms involved, are extremely rare. Through long-term, range-wide resightings of hundreds of colour-marked individuals from their first migration to adulthood and the application of state-space models, we quantify levels of juvenile and adult regional-scale movements and distances at different life stages across the whole non-breeding distribution range in a migratory shorebird, the Black-tailed Godwit (Limosa limosa islandica). We show that the probability of individuals changing non-breeding regions (seven historical wintering regions spanning the Western Europe range) at all ages is very low (mean movement probability = 10.9% from first to subsequent winter, and 8.3% from first adult winter to later winters). Movement between regions was also low between autumn and winter of the same year for both juveniles (mean movement probability = 17.0%) and adults (10.4%). The great majority of non-breeding movements from the first autumn to adulthood were within regions and less than 100 km. The scarcity of regional-scale non-breeding movements from the first autumn to adulthood means that the factors influencing where juveniles settle will be key determinants of non-breeding distributions and of the rate and direction of changes in distributions.

Nuclear quantum memory for hard x-ray photon wave packets.

Optical quantum memories are key elements in modern quantum technologies to reliably store and retrieve quantum information. At present, they are conceptually limited to the optical wavelength regime. Recent advancements in x-ray quantum optics render an extension of optical quantum memory protocols to ultrashort wavelengths possible, thereby establishing quantum photonics at x-ray energies. Here, we introduce an x-ray quantum memory protocol that utilizes mechanically driven nuclear resonant 57Fe absorbers to form a comb structure in the nuclear absorption spectrum by using the Doppler effect. This room-temperature nuclear frequency comb enables us to control the waveform of x-ray photon wave packets to a high level of accuracy and fidelity using solely mechanical motions. This tunable, robust, and highly flexible system offers a versatile platform for a compact solid-state quantum memory at room temperature for hard x-rays.

Impact of simulation fidelity on identifying swing-over hazards in virtual environments for novice crane operators

Crane-related accidents often occur due to failure in timely and accurate hazard identification. Virtual Reality (VR) technology facilitates immersive environments for improving hazard identification without real-world consequences. However, the impact of simulation fidelity on hazard identification performance, particularly for novice crane operators, remains underexplored. This paper examines how simulation fidelity and task complexity affect hazard identification performance among novice operators through a VR experiment with two fidelity and two complexity levels. Findings demonstrate high-fidelity (HF) scenarios significantly improve hazard identification success rate and speed for swing-over hazards but extend task completion times. Task complexity modulates simulation fidelity's impact on task completion, with significant differences between fidelity levels observed only under low complexity. These insights inform the design and calibration of simulation fidelity in VR platforms, potentially improving training efficacy and crane lifting safety for novice operators.

Numerical simulation of involute-plate research reactor flow behavior using RANS, LES and DNS

This paper investigates the flow behavior of involute-plate research reactors by performing Reynolds-Averaged Navier Stokes simulation (RANS), Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) of the channel flow between fuel plates. By modeling turbulence with different numerical approaches, this study provides data with three levels of fidelity. For the RANS simulation, three widely used turbulence models, i.e., k-ε, k-ω, Reynolds Stress Turbulence model (RST) are applied by using the commercial CFD code STAR-CCM + . For LES and DNS, the open-source CFD code, Nek5000, is used given its outstanding scalability on High Performance Computer (HPC) and high-order technique. The results from RANS simulations are compared with that from LES and DNS for benchmarking. Both macroscale parameters and turbulence statistics, such as velocity magnitude, lateral velocity and turbulence kinetic energy, are presented and analyzed.The results from RANS simulation achieve good agreement with LES and DNS on velocity and turbulence kinetic energy prediction. The RST turbulence model predicts the most similar flow pattern of lateral velocity as compared to LES and DNS. The Lambda-2 (λ2) criterion with a reasonable threshold is used to demonstrate the instantaneous vortices distribution in the involute channel from both LES and DNS calculation. The DNS simulation captures more detailed turbulence especially near the corner, which explains the discrepancy between LES and DNS results near the corner. The normalized RMS error are defined and calculated to assess the performance of those turbulence models. The RST model captures the anisotropic feature of turbulence, which enable it to outperform other turbulence models for predicting the flow behavior in an involute channel. Although some discrepancies are found between LES and DNS results in the corner, the overall deviations between LES and DNS are found to be small. Given that the computational cost of DNS calculation is an order of magnitude higher, using LES data for benchmarking RANS model is a cost-effective approach.

Generative adversarial networks for multi-fidelity matrix completion with massive missing entries

Multi-fidelity matrices refer to a pair of data matrices, whose entries are the measurements of a specific physical quantity at different fidelity levels w.r.t. two environment variables arranged as the row and the column of matrices, respectively. Although a high-fidelity (HF) matrix can provide an accurate description of the physical quantity, it usually has a number of missing entries because of high acquisition cost. In contrast, its low-fidelity (LF) counterpart is complete but the accuracy of its entries cannot meet the actual requirement. Multi-fidelity matrix completion (MMC) aims to complete the missing entries of the HF matrix by fusing the LF matrix and the available entries of the HF matrix. In this paper, we develop a generative adversarial network for multi-fidelity matrix completion (GAN-MMC), which is composed of a generator and a discriminator. The generator input is the LF matrix and its output is the approximation of the complete HF matrix. The discriminator identifies whether the HF approximation completed by the generator accords with the real HF data distribution. A matrix data augmentation method is proposed to enrich the samples for training GAN-MMC. Moreover, a correlation coefficient-based term is also imposed into the generative loss in order to make the generator capture the relatedness information between the available HF entries and the corresponding LF entries. The numerical experiments show that the proposed model can effectively complete the HF matrix with massive missing entries.

A Framework for Developing Systematic Testbeds for Multi-Fidelity Optimization Techniques

Abstract Multi-fidelity (MF) models abound in simulation-based engineering fields. Many MF strategies have been proposed to improve the efficiency in engineering processes, especially in design optimization. When it comes to assessing the performance of MF optimization techniques, existing practice usually relies on test cases involving contrived MF models of seemingly random math functions, due to limited access to real-world MF models. While it is acceptable to use contrived MF models, these models are often manually written up rather than created in a systematic manner. This gives rise to the potential pitfall that the test MF models may be not representative of general scenarios. We propose a framework to generate test MF models systematically and characterize tested MF optimization methods' performances comprehensively. In our framework, the MF models are generated based on given high-fidelity (HF) model and come with two parameters to control their fidelity levels and allow model randomization. In our testing process, MF case problems are systematically formulated using our model creation method. Running the given MF optimization technique on these problems produces what we call “savings curve” that characterizes the method's performance similarly to how ROC curves characterize machine learning classifiers. Our test results also allow plotting “optimality curves” that serve similar functionality to savings curves in certain types of problems. The flexibility of our MF model creation facilitates the development of testing processes for general MF problem scenarios, and our framework can be easily extended to other MF applications than optimization.

Implementation of training and coaching to improve goal planning and family engagement in early childhood home visiting

Early childhood home visiting services aim to support child and family health and development, yet challenges engaging and retaining families persist. One promising strategy to improve family engagement involves enhancing the capacity of the home visiting workforce to partner with families to set and implement service goals that align with families’ needs and preferences. This study examined implementation of ‘Goal Plan Strategies’ (GPS), a training and coaching innovation designed to improve family-centered goal planning. Aims were to evaluate (1) the extent to which training and eight sessions of audio-based coaching were implemented with fidelity, (2) whether coaching fidelity was associated with home visitors’ perceptions of coaching and confidence using goal planning strategies, and (3) whether GPS was perceived as feasible, acceptable, and useful by coaches and home visitors. Home visitors (N = 22) and coaches (N = 10) took part in GPS and completed surveys, session logs, and interviews. Training and coaching sessions were observed and coded for fidelity (e.g., dose, adherence, quality). GPS was implemented with high levels of fidelity. Findings support feasibility, acceptability, and usefulness of GPS, overall, and suggest the need to examine coaching dosage and overlap with other implementation supports more closely. Findings highlight the benefits of audio-based coaching, mentoring of coaches, and observational measures of implementation fidelity.

Participatory inventory of plant-based ethnoveterinary medicine used to control internal parasites of goats in the Ngamiland region of Botswana

The effect of internal parasites on goats is of major concern, particularly in the tropics where there are limited resources and financially constrained farmers. The ethnic knowledge on controlling and treating internal parasites is of value to sustainable and organic product production which are essential for long-term prosperity of goat farmers. Using participatory approaches, this study identified and documented popular plants used by farmers to control internal parasites in goats in the Ngamiland region of Botswana. A total of 120 communal farmers were interviewed using a structured questionnaire consisting of demographic parameters and ethnobotanical information. The data was analysed using Statistical Package for the Social Sciences (SPSS) and the most popular plants were identified using fidelity levels (FL). Most of the informants were secondary educated males of 50 to 60 years and relied on livestock for their livelihoods. It was recorded that a total of 13 local plant species from 11 families are commonly used in the treatment of intestinal parasites in goats. The plant family Fabaceae (27.3 %) was common among the farmers. The roots (76.9 %) were the most frequently used plant part while decoction (46 %) was popular in medicinal preparations. The prepared plant-based medicine is often administered orally. The highest FL values were found for the following plant species; Terminalia sericea, Gnidia capitata and Aloe ferox, scoring 89.3 %, 88.0 % and 86.3 %, respectively. The safety and efficacy of these plants should be validated to inform strategic utilization and alternative cost-effective strategy in managing internal parasites in Botswana.

Teleportation of Multiparticle Entangled States

This paper investigates the theoretical basis and actual methods for teleporting entangled states using multiple particles. Beginning with Bennett et al.'s pioneering work in 1993, which established the notion of quantum teleportation. In this paper, I will discuss the quantum teleportation, quantum entanglement, Entanglement schemes such as Entanglement swapping scheme and will find out the rate of successful teleportation in terms of fidelity. Additionally discussed is minimal assured fidelity, which is the lowest level of fidelity for any unknown quantum information about the states[1]. Additionally, the success rate of teleporting a superposition of odd and even coherent states is demonstrated to be able to be raised from 50% to almost 99%[2]. We discover that by just isolating the vacuum state from the even state, it can virtually transport, divert, and exchange. Key Words: Entanglement, Quantum teleportation, Entanglement swapping scheme, Bell state measurement (BSM), Quantum information.

In-Water Photo Identification, Site Fidelity, and Seasonal Presence of Harbor Seals (Phoca vitulina richardii) in Burrows Pass, Fidalgo Island, Washington

Little is known about the in-water behavior and site fidelity of harbor seals (Phoca vitulina richardii), as most photo-identification (photo-ID) studies are typically conducted while they are hauled-out on land. We investigated in-water site fidelity rates and seasonal presence in Burrows Pass, Washington, using photographs collected during a long-term photo-ID and behavioral study from January 2015 through November 2019. There was a minimum of 161 individuals and a maximum of 286 individual harbor seals using Burrows Pass. Harbor seals were present in all seasons, with the lowest sighting rates during summer. Individuals were more likely to be sighted/re-sighted in fall and spring. There was large variations in the level and seasonality of site fidelity among individuals. The majority of seals (69.62%) were seen only once, but 22.69% showed low to moderate site fidelity (2–5 sightings) and 7.69% showed strong site fidelity (≥6 sightings) over seasons and across years. These seasonal variations were likely due to foraging, life history, and individual behavioral variabilities. Studies like this provide necessary information about harbor seal in-water site fidelity and behavior, which are less well known but vitally important in harbor seal management and conservation.

Two‐step simulation combining large eddy simulation and population balance Monte Carlo for nanoparticle synthesis in flame spray pyrolysis

AbstractA two‐step simulation approach combining large eddy simulation (LES) for turbulent flame and multidimensional/multivariate population balance Monte Carlo (PBMC) for nanoparticle dynamics is proposed to explore the detailed flame fields and particle dynamics of zirconia nanoparticles in a flame spray pyrolysis (FSP) reactor. The turbulent combustion of gas and spray are simulated by the nonlinear LES‐partially stirred reactor model. Thereafter, the differentially weighted PBMC is applied for the first time to describe the spatially resolved formation and growth of nanoparticles using the flame fields derived from LES as an input. An efficient submodel for particle spatial transport in multidimensional grids is adopted and tested to account for the effect of thermophoresis, convective and diffusion of the nanoparticles. This methodology is successfully applied to an FSP case, demonstrating a reliable level of fidelity when compared to experiments and other model. Simulation results are discussed, in particular the flame fields and the size, morphology, as well as polydisperse primary particle and aggregates size distribution.

Multi-fidelity reinforcement learning with control variates

In many computational science and engineering applications, the output of a system of interest corresponding to a given input can be queried at different levels of fidelity with different costs. Typically, low-fidelity data is cheap and abundant, while high-fidelity data is expensive and scarce. In this work we study the reinforcement learning (RL) problem in the presence of multiple environments with different levels of fidelity for a given control task. We focus on improving the RL agent’s performance with multifidelity data. Specifically, a multifidelity estimator that exploits the cross-correlations between the low- and high-fidelity returns is proposed to reduce the variance in the estimation of the state–action value function. The proposed estimator, which is based on the method of control variates, is used to design a multifidelity Monte Carlo RL (MFMCRL) algorithm that improves the learning of the agent in the high-fidelity environment. The impacts of variance reduction on policy evaluation and policy improvement are theoretically analyzed by using probability bounds. Our theoretical analysis and numerical experiments demonstrate that for a finite budget of high-fidelity data samples, our proposed MFMCRL agent attains superior performance compared with that of a standard RL agent that uses only the high-fidelity environment data for learning the optimal policy. 22This article is an extended version of a paper originally presented at the ESANN (European Symposium on Artificial Neural Networks) conference in 2023.