Default Version Research Articles

Explainable SE-MobileNet for Pneumonia detection integrated with robustness assessment using adversarial examples

Pneumonia is a detrimental disease, especially for children, which is caused due to bacterial infection. X-ray images are frequently observed manually to find out the existence of pneumonia in a patient’s body. However, diagnosing pneumonia using X-ray images through manual observation by different health professionals may lead to different conclusions. Thus, an efficient autonomic system is required to diagnose pneumonia from X-ray images, and deep learning techniques, such as CNN-based approaches are frequently used to create such autonomy. To ease the process of pneumonia diagnosis, in this study, we have proposed the SE-MobileNet approach. We compared the performance of the proposed SE-MobileNet with the default version of MobileNetV2 integrated with transfer learning. Using the publicly available Kaggle dataset, it is observed that SE-MobileNet obtained 97.4% accuracy on a select test set against the 96.4% accuracy of MobileNetV2, and in 10-fold cross-validation, SE-MobileNet achieved an average of 95.92% accuracy against the 92.35% accuracy of MobileNetV2. Further comparison analysis proves that the SE-MobileNet model not only performs much better than the vanilla MobileNetV2 but also performs competitively against the literature. In addition, robustness evaluation has been introduced in this study where Fast Gradient Sign Method (FGSM) is performed to generate adversarial images. It is found that in robustness evaluation, SE-MobileNet also performs better compared to MobileNetV2. Finally, to validate the appropriateness of the learning of the model, Explainable AI (XAI) based techniques have been employed.

Keeping Up With ChatGPT: Evaluating Its Recognition and Interpretation of Nuclear Medicine Images.

The latest iteration of GPT4 (generative pretrained transformer) is a large multimodal model that can integrate both text and image input, but its performance with medical images has not been systematically evaluated. We studied whether ChatGPT with GPT-4V(ision) can recognize images from common nuclear medicine examinations and interpret them. Fifteen representative images (scintigraphy, 11; PET, 4) were submitted to ChatGPT with GPT-4V(ision), both in its Default and "Advanced Data Analysis (beta)" version. ChatGPT was asked to name the type of examination and tracer, explain the findings and whether there are abnormalities. ChatGPT should also mark anatomical structures or pathological findings. The appropriateness of the responses was rated by 3 nuclear medicine physicians. The Default version identified the examination and the tracer correctly in the majority of the 15 cases (60% or 53%) and gave an "appropriate" description of the findings or abnormalities in 47% or 33% of cases, respectively. The Default version cannot manipulate images. "Advanced Data Analysis (beta)" failed in all tasks in >90% of cases. A "major" or "incompatible" inconsistency between 3 trials of the same prompt was observed in 73% (Default version) or 87% of cases ("Advanced Data Analysis (beta)" version). Although GPT-4V(ision) demonstrates preliminary capabilities in analyzing nuclear medicine images, it exhibits significant limitations, particularly in its reliability (ie, correctness, predictability, and consistency).

Enhancing cladding mechanical modelling during DBA/LOCA accidents with FRAPTRAN: The TUmech one-dimensional model

The analysis of the cladding mechanical performance is essential for predicting rod failure during the heat-up phase of a loss-of-coolant accident (LOCA). Previous modelling and simulation exercises have shown that there is room for improvement in the accuracy of fuel performance simulations under such conditions. In this work, TUmech, a simplified standalone version of the mechanical model in TRANSURANUS, has been developed and coupled with FRAPTRAN-2.0. After highlighting the most significant differences between both approaches, particularly in their mechanical treatment and material property correlations, simulations have been performed for IFA-650.2 (fresh fuel) and IFA-650.10 (high burnup fuel) of the Halden LOCA test series. The results obtained from both FRAPTRAN versions have been compared with experimental data and benchmarked against TRANSURANUS estimates, focusing on rod inner gas pressure, post-test cladding outer diameter, and burst time. In both scenarios, FRAPTRAN-TUmech has proven to enhance the accuracy in predicting these three variables compared to the FRAPTRAN default version. However, further validation is required before drawing definitive conclusions based on the encouraging results.

JWST’s PEARLS: Improved Flux Calibration for NIRCam

The Prime Extragalactic Areas for Reionization and Lensing Science, a James Webb Space Telescope (JWST) GTO program, obtained a set of unique NIRCam observations that have enabled us to significantly improve the default photometric calibration across both NIRCam modules. The observations consisted of three epochs of 4-band (F150W, F200W, F356W, and F444W) NIRCam imaging in the Spitzer IRAC Dark Field (IDF). The three epochs were six months apart and spanned the full duration of Cycle 1. As the IDF is in the JWST continuous viewing zone, we were able to design the observations such that the two modules of NIRCam, modules A and B, were flipped by 180° and completely overlapped each other’s footprints in alternate epochs. We were therefore able to directly compare the photometry of the same objects observed with different modules and detectors, and we found significant photometric residuals up to ∼0.05 mag in some detectors and filters, for the default version of the calibration files that we used (jwst_1039.pmap). Moreover, there are multiplicative gradients present in the data obtained in the two long-wavelength bands. The problem is less severe in the data reduced using the latest pmap (jwst_1130.pmap as of 2023 September), but it is still present, and is non-negligible. We provide a recipe to correct for this systematic effect to bring the two modules onto a more consistent calibration, to a photometric precision better than ∼0.02 mag.

Carousel Greedy Algorithms for Feature Selection in Linear Regression

The carousel greedy algorithm (CG) was proposed several years ago as a generalized greedy algorithm. In this paper, we implement CG to solve linear regression problems with a cardinality constraint on the number of features. More specifically, we introduce a default version of CG that has several novel features. We compare its performance against stepwise regression and more sophisticated approaches using integer programming, and the results are encouraging. For example, CG consistently outperforms stepwise regression (from our preliminary experiments, we see that CG improves upon stepwise regression in 10 of 12 cases), but it is still computationally inexpensive. Furthermore, we show that the approach is applicable to several more general feature selection problems.

An attempt to decrease social desirability bias: The effect of cheap talk mitigation on internal and external validity of discrete choice experiments

Discrete choice experiments (DCEs) have been used extensively to elicit preferences. However, the hypothetical nature of choices induces socially desirable behaviour and endangers internal and external validity of DCEs. This studyexperimentally investigated social desirability bias in DCEs and whether it can be mitigated using the cheap talk mitigation method in the context of food choices. Respondents (N = 1027) were randomly allocated to one of four questionnaire versions: default without manipulation, priming socially desirable behaviour, cheap talk mitigation, or both. The effect on internal validity was assessed by comparing respondent-reported characteristics, DCE results, and prediction accuracy for a holdout task between questionnaire versions. The effect on external validity was assessed by comparing stated and revealed preferences. Social desirability bias, if present, was hardly affected by cheap talk mitigation. Respondent-reported characteristics, DCE results and prediction accuracy for the holdout task and actual food choice did not strongly differ between questionnaire versions.Prediction accuracy for the holdout task was lowest in the default version. Prediction accuracy for actual food choice was slightly better among respondents in the versions that were exposed to cheap talk mitigation. Social desirability bias was hard to detect and mitigate in this study, potentially due to limited social desirability, the effectiveness of the cheap talk mitigation method, and other sources of hypothetical bias. The differences in prediction accuracy indicates that cheap talk mitigation slightly improved external validity at minimum cost to internal validity. Recommendations for future research are provided.

Fire weather index data under historical and shared socioeconomic pathway projections in the 6th phase of the Coupled Model Intercomparison Project from 1850 to 2100

Abstract. Human-induced climate change is increasing the incidence of fire events and associated impacts on livelihood, biodiversity, and nature across the world. Understanding current and projected fire activity together with its impacts on ecosystems is crucial for evaluating future risks and taking actions to prevent such devastating events. Here we focus on fire weather as a key driver of fire activity. Fire weather products that have a global homogenous distribution in time and space provide many advantages to advance fire science and evaluate future risks. Therefore, in this study we calculate and provide for the first time the Canadian Fire Weather Index (FWI) with all available simulations of the 6th phase of the Coupled Model Intercomparison Project (CMIP6). Furthermore, we expand its regional applicability by combining improvements to the original algorithm for the FWI from several packages. A sensitivity analysis of the default version versus our improved version shows significant differences in the final FWI. With the improved version, we calculate the FWI using average relative humidity in one case and minimum relative humidity in another case. We provide the data for both cases while recommending the one with minimum relative humidity for studies focused on actual FWI values and the one with average relative humidity for studies requiring larger ensembles. The following four annual indicators, (i) maximum value of the FWI (fwixx), (ii) number of days with extreme fire weather (fwixd), (iii) length of the fire season (fwils), and (iv) seasonal average of the FWI (fwisa), are made available and are illustrated here. We find that, at a global warming level of 3 ∘C, the mean fire weather would increase on average by at least 66 % in duration and frequency, while associated 1-in-10-year events would approximately triple in duration and increase by at least 31 % in intensity. Ultimately, this new fire weather dataset provides a large ensemble of simulations to understand the potential impacts of climate change spanning a range of shared socioeconomic narratives with their radiative forcing trajectories over 1850–2100 at annual and 2.5∘ × 2.5∘ resolutions. The produced full global dataset is a freely available resource at https://doi.org/10.3929/ethz-b-000583391 (Quilcaille and Batibeniz, 2022) for fire danger studies and beyond, which highlights the need to reduce greenhouse gas emissions for reducing fire impacts.

Exploring critical fluctuation phenomenon according to net-proton multiplicity information entropy in AMPT model

Information entropy can be used for investigating the critical fluctuation phenomenon by describing multiplicity fluctuations in heavy-ion collisions. In this study, a multiphase transport (AMPT) model in the default version (AMPT-DEF) and with the string melting (AMPT-SM) were used as event generators to simulate the Au+Au reaction at different incident energies. Owing to the influence of the distribution bin number in extracting its information entropy, two methods, i.e., data binning and weighting factors, were proposed to correct the information entropy analysis. The comparison between information entropy and kurtosis also reflects the necessity to correct the information entropy analysis. The corrected information entropy of net-proton multiplicity was analyzed in central Au+Au collisions at sNN= 4.7, 6.4, 7.7, 8.8, 11.5, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 140, 170, and 200 GeV with the impact parameter b=3 fm. The results indicated that the fluctuations of net-proton multiplicity were essentially identical at collision energies ranging from sNN= 4.7 GeV to sNN= 200 GeV in the same model. No critical fluctuation phenomenon was found in the central Au+Au collisions simulated using the AMPT models, similar to the case of the AMPT model, which does not have. Comparing the information entropy of AMPT-SM to AMPT-DEF, the effect of partonic phase leads to a bigger fluctuation, indicating that the corrected information entropy can be a useful probe to explore the phase transition. The mid-rapidity net-proton multiplicity information entropy was also calculated, and the results led to the same conclusion. The corrected information entropy may be useful for investigating critical fluctuations, which is important for exploring critical endpoint.

How more sophisticated leaf biomass simulations can increase the realism of modelled animal populations

Animal biodiversity, and its key roles in ecosystem state and functioning, is facing critical challenges in the wake of anthropogenic activities. It is urgently necessary to improve understanding of the interconnections between animals and the vegetation within ecosystems. Process-based modelling has shown to be a mighty tool in making assessments on ecological processes. We assess the effect of different vegetation models on simulated animal biodiversity by replacing the vegetation module within Madingley, a multi-trophic model of functional diversity with LPJ-GUESS, a dynamic global vegetation model. We compare the output metrics of the model system to Madingley’s default version for four ecosystem types around the globe and analyse whether the realism of the simulation results increased as a result of the coupling between Madingley and LPJ-GUESS. Simulated animal populations react to the coupling by shifting towards smaller individuals with a higher abundance. General shifts in body mass and animal distributions can be traced back to ecological processes, allowing in-depth analysis of heterotrophic responses to changes in leaf biomass. We also derive power-law relationships for herbivory to NPP and herbivore biomass to NPP and conclude that the coupled model system simulates animal populations that follow reasonable power-laws which are similar to power-laws derived from empirical data. Our results indicate that developing process-based model systems is a viable way to assess multi-trophic interconnections between animal populations and the ecosystems vegetation.

Simulating cold shear flows on a moving mesh

ABSTRACT Rotationally supported, cold, gaseous discs are ubiquitous in astrophysics and appear in a diverse set of systems, such as protoplanetary discs, accretion discs around black holes, or large spiral galaxies. Capturing the gas dynamics accurately in these systems is challenging in numerical simulations due to the low sound speed compared to the bulk velocity of the gas, the resolution limitations of full disc models, and the fact that numerical noise can easily source spurious growth of fluid instabilities if not suppressed sufficiently well, negatively interfering with real physical instabilities present in such discs (like the magnetorotational instability). Here, we implement the so-called shearing-box approximation in the moving-mesh code arepo in order to facilitate achieving high resolution in local regions of differentially rotating discs and to address these problems. While our new approach offers manifest translational invariance across the shearing-box boundaries and offers continuous local adaptivity, we demonstrate that the unstructured mesh of arepo introduces unwanted levels of ‘grid-noise’ in the default version of the code. We show that this can be rectified by high-order integrations of the flux over mesh boundaries. With our new techniques we obtain highly accurate results for shearing-box calculations of the magnetorotational instability that are superior to other Lagrangian techniques. These improvements are also of value for other applications of the code that feature strong shear flows.

Modelling Inhomogeneous Geodata Quality in a Dataset’s Metadata

Abstract. Extensive data quality descriptions as a vital part of a dataset’s metadata are widely accepted, albeit their provision in a formalized manner is often lacking. This is due to a number of problems that are frequently encountered by geodata producing scientists. As one of these problems, we identified missing, unknown or unused options to model inhomogeneity of data quality across space, time, and theme in a dataset’s metadata. Detailed information of inhomogeneous geodata quality beyond dataset-wide statistical measures (variance, min, max, etc.) is often only described in dataset accompanying papers or quality reports. These text-based approaches prevent precise querying and hinder the development of advanced data discovery tools that could make valuable use of inhomogeneous data quality information. We propose a profile for the data quality vocabulary (DQV) that allows to model inhomogeneous geodata quality. Considering established vocabularies typically used to describe geographic metadata, as well as ensuring compatibility with the default version of DQV, enhances the usability and thus, minimizes the effort for data producers to provide formalized descriptions of inhomogeneous data quality.

SPLICE:a synthetic paid loss and incurred cost experience simulator

AbstractIn this paper, we first introduce a simulator of cases estimates ofincurred lossescalledSPLICE(SyntheticPaidLoss andIncurredCostExperience). In three modules, case estimates are simulated in continuous time, and a record is output for each individual claim. Revisions for the case estimates are also simulated as a sequence over the lifetime of the claim in a number of different situations. Furthermore, some dependencies in relation to case estimates of incurred losses are incorporated, particularly recognising certain properties of case estimates that are found in practice. For example, the magnitude of revisions depends on ultimate claim size, as does the distribution of the revisions over time. Some of these revisions occur in response to occurrence of claim payments, and soSPLICErequires input of simulated per-claim payment histories. The claim data can be summarised by accident and payment “periods” whose duration is an arbitrary choice (e.g. month, quarter, etc.) available to the user.SPLICEis built on an existing simulator of individual claim experience calledSynthETIC(introduced in Avanziet al. 2021b,Insurance: Mathematics and Economics,100, 296–308), which offers flexible modelling of occurrence, notification, as well as the timing and magnitude of individual partialpayments. This is in contrast with theincurred losses, which constitute the additional contribution ofSPLICE. The inclusion of incurred loss estimates provides a facility that almost no other simulators do.SPLICEis is a fully documented R package that is publicly available and open source (on CRAN).SPLICE, combined withSynthETIC, provides 11 modules (occurrence, notification, etc.), any one or more of which may be re-designed according to the user’s requirements. It comes with a default version that is loosely calibrated to resemble a specific (but anonymous) Auto Bodily Injury portfolio, as well as data generation functionality that outputs alternative data sets under a range of hypothetical scenarios differing in complexity. The general structure is suitable for most lines of business, with some reparameterisation.

Improved runoff simulations for a highly varying soil depth and complex terrain watershed in the Loess Plateau with the Community Land Model version 5

Abstract. This study aimed to improve runoff simulations and explore deep soil hydrological processes for a watershed in the center of the Loess Plateau (LP), China. This watershed, the Wuding River Basin (WRB), has very complex topography, with soil depths ranging from 0 to 197 m. The hydrological model used for our simulations was Community Land Model version 5 (CLM5) developed by the National Center for Atmospheric Research. Actual soil depths and river channels were incorporated into CLM5 to realistically represent the physical features of the WRB. Through sensitivity tests, CLM5 with 150 soil layers with the observed variable soil depths produced the most reasonable results and was adopted for this study. Our results showed that CLM5 with actual soil depths significantly suppressed unrealistic variations of the simulated subsurface runoff when compared to the default simulations. In addition, when compared with the default version with 20 soil layers, CLM5 with 150 soil layers slightly improved runoff simulations but generated simulations with much smoother vertical water flows that were consistent with the uniform distribution of soil textures in our study watershed. The runoff simulations were further improved by the addition of river channels to CLM5, where the seasonal variability of the simulated runoff was reasonably captured. Moreover, the magnitude of the simulated runoff remarkably decreased with increased soil evaporation by lowering the soil water content threshold, which triggers surface resistance. The lowered threshold was consistent with the loess soil, which has a high sand component. Such soils often generate stronger soil evaporation than soils dominated by clay. Finally, with the above changes in CLM5, the simulated total runoff matched very closely with observations. When compared with those for the default runoff simulations, the correlation coefficient, root mean square error, and Nash–Sutcliffe coefficient for the improved simulations changed dramatically from 0.02, 10.37 mm, and −12.34 to 0.62, 1.8 mm, and 0.61. The results in this study provide strong physical insight for further investigation of hydrological processes in complex terrain with deep soils.

Automated detection of photovoltaic cleaning events: A performance comparison of techniques as applied to a broad set of labeled photovoltaic data sets

AbstractExtracting accurate soiling loss information from photovoltaic (PV) production data first requires segmenting the time series data per natural or manually occurring cleaning events. Maintenance logs are often incomplete, rain data are often unavailable, and the debate on rain thresholds for cleaning and dew or wind cleanings is still ongoing. The present work aims to overtake these issues by improving automated methods to detect these cleaning events and therefore improve extraction of soiling loss information. Time series power production data from 22 PV inverters were labeled for natural or manually occurring cleaning events. The data sets were carefully selected to include varying degrees of soiling, cleaning events, and noise. Several algorithms, including filtering logic and change point detection, were examined for efficacy at detecting the labeled cleanings. All the methods introduced except for changepoint detection showed significant improvement at detecting the labeled cleaning events per the mean F1 score. Furthermore, the highest performing cleaning detection algorithm achieved an absolute increase in the mean F1 score of 43% over the default version of the RdTools stochastic rate and recovery (SRR) algorithm. The highest performing algorithm included irradiance filtering and a cleaning detection threshold, adjusted based on the 40‐day centered rolling median of the absolute day‐to‐day deviations in the daily performance index (PI). These improvements are promising as cleaning detection is an essential step in the automated analysis of PV soiling.

Models under which random forests perform badly; consequences for applications

We give examples of data-generating models under which Breiman’s random forest may be extremely slow to converge to the optimal predictor or even fail to be consistent. The evidence provided for these properties is based on mostly intuitive arguments, similar to those used earlier with simpler examples, and on numerical experiments. Although one can always choose models under which random forests perform very badly, we show that simple methods based on statistics of ‘variable use’ and ‘variable importance’ can often be used to construct a much better predictor based on a ‘many-armed’ random forest obtained by forcing initial splits on variables which the default version of the algorithm tends to ignore.

Alvis Approach to Modeling and Verification of Real-Time Systems Running on Single-Processor Environment

Alvis is a formal modelling language developed primarily for modelling concurrent systems including real-time systems. The prepared model is compiled and the resulting runnable model allows, e.g., to generate the state space and to verify the model using model checking techniques. The default version assumes that each system component (agent) executes its computations in parallel and that the hardware platform is not a shared resource. This paper describes an alternative approach in which agents remain concurrent, but compete for access to a single shared processor. This situation significantly affects the correctness of the system, in particular from the perspective of the timing properties of real-time systems. The key element of the presented solution is the fact that changing the runtime environment comes down to changing compiler options and does not require changes to the original model. This paper presents a method of designing Alvis language models for a single-processor hardware platform. The presented concepts are illustrated with examples.

Outdoor performance of the black globe temperature sensor on a hot and humid tropical region

ABSTRACT A crucial variable to evaluate thermal comfort is the mean radiant temperature (Tmrt). In this paper we evaluate the performance of the 150 mm black globe thermometer to provide reliable Tmrt values for outdoor settings in Singapore. Accurate Tmrt values are calculated by the method of integral radiation measurements. Based on these, the mean convection coefficient of the black globe has been re-calibrated. Results show an improvement in the estimation of Tmrt with the new coefficient in comparison with the default version suggested in ISO7726:1998. Increasing the averaging periods of the measured variables improved the performance of the derived mean convective coefficients to estimate Tmrt. During clear skies day and for 10-min averaged data, RMSE for Tmrt reduce to 3.9 °C (7.4 °C for ISO7726:1998 coefficient) with an overestimation on high incoming solar radiation periods and an underestimation during the morning and evening (low solar elevation). During overcast dry conditions an underestimation of Tmrt is also expected which is higher in the rain/wet periods. The mean convective coefficient presented in this work can provide improved estimations of Tmrt relevant for outdoor thermal comfort studies in hot and humid tropical climates like Singapore.

Impact of Stochastic Entrainment in the NCAR CAM Deep Convection Parameterization on the Simulation of South Asian Summer Monsoon

Model simulations are highly sensitive to the formulation of the atmospheric mixing process or entrainment in the deep convective parameterizations used in their atmospheric component. In this paper, we have implemented stochastic entrainment in the deep convection scheme of NCAR CAM5 and analyzed the improvements in model simulation, focusing on the South Asian Summer Monsoon (SASM), as compared to the deterministic entrainment formulation in the default version of the model. Simulations using stochastic entrainment (StochCAM5) outperformed default model simulations (DefCAM5), as inferred from multiple metrics associated with the SASM. StochCAM5 significantly alleviated some of the longstanding SASM biases seen in DefCAM5, such as precipitation pattern and magnitude over the Arabian Sea and western Equatorial Indian ocean, early monsoon withdrawal, and the overestimation in the frequency of light precipitation and the underestimation in the frequency of large-to-extreme precipitation. Related SASM dynamical and thermodynamical features, such as Somali Jet, low-level westerly winds, and meridional tropospheric temperature gradient (MTTG), are improved in StochCAM5. Further, the simulation of monsoon intra-seasonal oscillation (MISO), Madden Julian Oscillation (MJO), and equatorial Kelvin waves are improved in StochCAM5. Many essential climate variables, such as shortwave and longwave cloud forcing, cloud cover, relative and specific humidity, and precipitable water, show significant improvement in StochCAM5.

SynthETIC: An individual insurance claim simulator with feature control

Recent years have seen rapid increase in the application of machine learning to insurance loss reserving. They yield most value when applied to large data sets, such as individual claims, or large claim triangles. In short, they are likely to be useful in the analysis of any data set whose volume is sufficient to obscure a naked-eye view of its features. Unfortunately, such large data sets are in short supply in the actuarial literature. Accordingly, one needs to turn to synthetic data. Although the ultimate objective of these methods is application to real data, the use of synthetic data containing features commonly observed in real data is also to be encouraged.While there are a number of claims simulators in existence, each valuable within its own context, the inclusion of a number of desirable (but complicated) data features requires further development. Accordingly, in this paper we review those desirable features, and propose a new simulator of individual claim experience called SynthETIC.Our simulator is publicly available, open source, and fills a gap in the non-life actuarial toolkit. The simulator specifically allows for desirable (but optionally complicated) data features typically occurring in practice, such as variations in rates of settlements and development patterns; as with superimposed inflation, and various discontinuities, and also enables various dependencies between variables. The user has full control of the mechanics of the evolution of an individual claim. As a result, the complexity of the data set generated (meaning the level of difficulty of analysis) may be dialed anywhere from extremely simple to extremely complex. The default version is parameterized so as to include a broad (though not numerically precise) resemblance to the major features of experience of a specific (but anonymous) Auto Bodily Injury portfolio, but the general structure is suitable for most lines of business, with some amendment of modules.

Logistic retainment interval dose exploration design for Phase I clinical trials of cytotoxic agents.

Phase I studies of a cytotoxic agent often aim to identify the dose that provides an investigator specified target dose-limiting toxicity (DLT) probability. In practice, an initial cohort receives a dose with a putative low DLT probability, and subsequent dosing follows by consecutively deciding whether to retain the current dose, escalate to the adjacent higher dose, or de-escalate to the adjacent lower dose. This article proposes a Phase I design derived using a Bayesian decision-theoretic approach to this sequential decision-making process. The design consecutively chooses the action that minimizes posterior expected loss where the loss reflects the distance on the log-odds scale between the target and the DLT probability of the dose that would be given to the next cohort under the corresponding action. A logistic model is assumed for the log odds of a DLT at the current dose with a weakly informative t-distribution prior centered at the target. The key design parameters are the pre-specified odds ratios for the DLT probabilities at the adjacent higher and lower doses. Dosing rules may be pre-tabulated, as these only depend on the outcomes at the current dose, which greatly facilitates implementation. The recommended default version of the proposed design improves dose selection relative to many established designs across a variety of scenarios.