More Like this

1. A simple physics-based improvement to the positive degree day model

   https://doi.org/10.1017/jog.2018.55  

   TSAI, VICTOR C.; RUAN, XIAOZHOU (August 2018, Journal of Glaciology)

   ABSTRACT Meltwater is important to understanding glacier health and dynamics. Since melt measurements are uncommon, ice ablation estimates are often based on models including the positive degree day (PDD) model. The PDD estimate is popular since it only requires air temperature as input, but suffers from the lack of physical motivation of an energy-balance model. We present a physics-based alternative to the PDD model that still only takes air/surface temperature as input. The model resembles the PDD model except accounting for time lags in ablation when cold ice needs to be warmed. The model is expressed as a differential equation with a single extra parameter related to the efficiency of heating a near-surface layer of ice. With zero thickness, the model reduces to the PDD model, providing a physical basis for the PDD model. Applying the model to data from Greenland, it improves modestly upon the PDD model, with the main improvement being better prediction of early season melting. This new model is a useful compromise, with some of the physics of more realistic models and the simplicity of a PDD model. The model should improve estimates of meltwater production and help constrain PDD parameters when empirical calibration is challenging. 

   more » « less
2. The Art and Science of Building a Computational Model to Understand Hemostasis

   https://doi.org/10.1055/s-0041-1722861  

   Leiderman, Karin; Sindi, Suzanne S.; Monroe, Dougald M.; Fogelson, Aaron L.; Neeves, Keith B. (February 2021, Seminars in Thrombosis and Hemostasis)

   null (Ed.)

   Abstract Computational models of various facets of hemostasis and thrombosis have increased substantially in the last decade. These models have the potential to make predictions that can uncover new mechanisms within the complex dynamics of thrombus formation. However, these predictions are only as good as the data and assumptions they are built upon, and therefore model building requires intimate coupling with experiments. The objective of this article is to guide the reader through how a computational model is built and how it can inform and be refined by experiments. This is accomplished by answering six questions facing the model builder: (1) Why make a model? (2) What kind of model should be built? (3) How is the model built? (4) Is the model a “good” model? (5) Do we believe the model? (6) Is the model useful? These questions are answered in the context of a model of thrombus formation that has been successfully applied to understanding the interplay between blood flow, platelet deposition, and coagulation and in identifying potential modifiers of thrombin generation in hemophilia A. 

   more » « less
3. Semiparametric imputation using conditional Gaussian mixture models under item nonresponse

   https://doi.org/10.1111/biom.13410  

   Lee, Danhyang; Kim, Jae Kwang (December 2020, Biometrics)

   Abstract Imputation is a popular technique for handling item nonresponse. Parametric imputation is based on a parametric model for imputation and is not robust against the failure of the imputation model. Nonparametric imputation is fully robust but is not applicable when the dimension of covariates is large due to the curse of dimensionality. Semiparametric imputation is another robust imputation based on a flexible model where the number of model parameters can increase with the sample size. In this paper, we propose a new semiparametric imputation based on a more flexible model assumption than the Gaussian mixture model. In the proposed mixture model, we assume a conditional Gaussian model for the study variable given the auxiliary variables, but the marginal distribution of the auxiliary variables is not necessarily Gaussian. The proposed mixture model is more flexible and achieves a better approximation than the Gaussian mixture models. The proposed method is applicable to high‐dimensional covariate problem by including a penalty function in the conditional log‐likelihood function. The proposed method is applied to the 2017 Korean Household Income and Expenditure Survey conducted by Statistics Korea. 

   more » « less
4. GAN-Based Domain Inference Attack

   https://doi.org/10.1609/AAAI.V37I12.26663  

   Gu, Yuechun; Chen, Keke (June 2023, Proceedings of the AAAI Conference on Artificial Intelligence)

   Model-based attacks can infer training data information from deep neural network models. These attacks heavily depend on the attacker's knowledge of the application domain, e.g., using it to determine the auxiliary data for model-inversion attacks. However, attackers may not know what the model is used for in practice. We propose a generative adversarial network (GAN) based method to explore likely or similar domains of a target model -- the model domain inference (MDI) attack. For a given target (classification) model, we assume that the attacker knows nothing but the input and output formats and can use the model to derive the prediction for any input in the desired form. Our basic idea is to use the target model to affect a GAN training process for a candidate domain's dataset that is easy to obtain. We find that the target model may distort the training procedure less if the domain is more similar to the target domain. We then measure the distortion level with the distance between GAN-generated datasets, which can be used to rank candidate domains for the target model. Our experiments show that the auxiliary dataset from an MDI top-ranked domain can effectively boost the result of model-inversion attacks. 

   more » « less
5. Discovery of interpretable structural model errors by combining Bayesian sparse regression and data assimilation: A chaotic Kuramoto–Sivashinsky test case

   https://doi.org/10.1063/5.0091282  

   Mojgani, Rambod; Chattopadhyay, Ashesh; Hassanzadeh, Pedram (June 2022, Chaos: An Interdisciplinary Journal of Nonlinear Science)

   Models of many engineering and natural systems are imperfect. The discrepancy between the mathematical representations of a true physical system and its imperfect model is called the model error. These model errors can lead to substantial differences between the numerical solutions of the model and the state of the system, particularly in those involving nonlinear, multi-scale phenomena. Thus, there is increasing interest in reducing model errors, particularly by leveraging the rapidly growing observational data to understand their physics and sources. Here, we introduce a framework named MEDIDA: Model Error Discovery with Interpretability and Data Assimilation. MEDIDA only requires a working numerical solver of the model and a small number of noise-free or noisy sporadic observations of the system. In MEDIDA, first, the model error is estimated from differences between the observed states and model-predicted states (the latter are obtained from a number of one-time-step numerical integrations from the previous observed states). If observations are noisy, a data assimilation technique, such as the ensemble Kalman filter, is employed to provide the analysis state of the system, which is then used to estimate the model error. Finally, an equation-discovery technique, here the relevance vector machine, a sparsity-promoting Bayesian method, is used to identify an interpretable, parsimonious, and closed-form representation of the model error. Using the chaotic Kuramoto–Sivashinsky system as the test case, we demonstrate the excellent performance of MEDIDA in discovering different types of structural/parametric model errors, representing different types of missing physics, using noise-free and noisy observations. 

   more » « less