Jian Du, Haochong Li, Qi Liao, Jun Shen, Jianqin Zheng, Yongtu Liang
The high-pressure transportation process of pipeline necessitates an accurate hydraulic transient simulation tool to prevent slack line flow and over-pressure, which can endanger pipeline operations. However, current numerical solution methods often face difficulties in balancing computational efficiency and accuracy. Additionally, few studies attempt to reform physics-informed learning architecture for pipeline transient simulation with magnitude different in outputs and imbalanced gradient in loss function. To address these challenges, a Knowledge-Inspired Hierarchical Physics-Informed Neural Network is proposed for hydraulic transient simulation of multi-product pipelines. The proposed model integrates governing equations, boundary conditions, and initial conditions into the training process to ensure consistency with physical laws. Furthermore, magnitude conversion of outputs and equivalent conversion of governing equations are implemented to enhance the training performance of the neural network. To further address the imbalanced gradient of multiple loss terms with fixed weights, a hierarchical training strategy is designed. Numerical simulations demonstrate that the proposed model outperforms state-of-the-art models and can still produce accurate simulation results under complex hydraulic transient conditions, with mean absolute percentage errors reduced by 87.8% and 92.7 % in pressure prediction. Thus, the proposed model can conduct accurate and effective hydraulic transient analysis, ensuring the safe operation of pipelines.
{"title":"A Knowledge-Inspired Hierarchical Physics-Informed Neural Network for Pipeline Hydraulic Transient Simulation","authors":"Jian Du, Haochong Li, Qi Liao, Jun Shen, Jianqin Zheng, Yongtu Liang","doi":"arxiv-2409.10911","DOIUrl":"https://doi.org/arxiv-2409.10911","url":null,"abstract":"The high-pressure transportation process of pipeline necessitates an accurate\u0000hydraulic transient simulation tool to prevent slack line flow and\u0000over-pressure, which can endanger pipeline operations. However, current\u0000numerical solution methods often face difficulties in balancing computational\u0000efficiency and accuracy. Additionally, few studies attempt to reform\u0000physics-informed learning architecture for pipeline transient simulation with\u0000magnitude different in outputs and imbalanced gradient in loss function. To\u0000address these challenges, a Knowledge-Inspired Hierarchical Physics-Informed\u0000Neural Network is proposed for hydraulic transient simulation of multi-product\u0000pipelines. The proposed model integrates governing equations, boundary\u0000conditions, and initial conditions into the training process to ensure\u0000consistency with physical laws. Furthermore, magnitude conversion of outputs\u0000and equivalent conversion of governing equations are implemented to enhance the\u0000training performance of the neural network. To further address the imbalanced\u0000gradient of multiple loss terms with fixed weights, a hierarchical training\u0000strategy is designed. Numerical simulations demonstrate that the proposed model\u0000outperforms state-of-the-art models and can still produce accurate simulation\u0000results under complex hydraulic transient conditions, with mean absolute\u0000percentage errors reduced by 87.8% and 92.7 % in pressure prediction. Thus,\u0000the proposed model can conduct accurate and effective hydraulic transient\u0000analysis, ensuring the safe operation of pipelines.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142249044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Shuaihao Zhang, Dong Wu, Sérgio D. N. Lourenço, Xiangyu Hu
Hourglass modes, characterized by zigzag particle and stress distributions, are a common numerical instability encountered when simulating solid materials with updated Lagrangian smoother particle hydrodynamics (ULSPH). While recent solutions have effectively addressed this issue in elastic materials using an essentially non-hourglass formulation, extending these solutions to plastic materials with more complex constitutive equations has proven challenging due to the need to express shear forces in the form of a velocity Laplacian. To address this, a generalized non-hourglass formulation is proposed within the ULSPH framework, suitable for both elastic and plastic materials. Specifically, a penalty force is introduced into the momentum equation to resolve the disparity between the linearly predicted and actual velocities of neighboring particle pairs, thereby mitigating the hourglass issue. The stability, convergence, and accuracy of the proposed method are validated through a series of classical elastic and plastic cases, with a dual-criterion time-stepping scheme to improve computational efficiency. The results show that the present method not only matches or even surpasses the performance of the recent essentially non-hourglass formulation in elastic cases but also performs well in plastic scenarios.
{"title":"A generalized non-hourglass updated Lagrangian formulation for SPH solid dynamics","authors":"Shuaihao Zhang, Dong Wu, Sérgio D. N. Lourenço, Xiangyu Hu","doi":"arxiv-2409.11474","DOIUrl":"https://doi.org/arxiv-2409.11474","url":null,"abstract":"Hourglass modes, characterized by zigzag particle and stress distributions,\u0000are a common numerical instability encountered when simulating solid materials\u0000with updated Lagrangian smoother particle hydrodynamics (ULSPH). While recent\u0000solutions have effectively addressed this issue in elastic materials using an\u0000essentially non-hourglass formulation, extending these solutions to plastic\u0000materials with more complex constitutive equations has proven challenging due\u0000to the need to express shear forces in the form of a velocity Laplacian. To\u0000address this, a generalized non-hourglass formulation is proposed within the\u0000ULSPH framework, suitable for both elastic and plastic materials. Specifically,\u0000a penalty force is introduced into the momentum equation to resolve the\u0000disparity between the linearly predicted and actual velocities of neighboring\u0000particle pairs, thereby mitigating the hourglass issue. The stability,\u0000convergence, and accuracy of the proposed method are validated through a series\u0000of classical elastic and plastic cases, with a dual-criterion time-stepping\u0000scheme to improve computational efficiency. The results show that the present\u0000method not only matches or even surpasses the performance of the recent\u0000essentially non-hourglass formulation in elastic cases but also performs well\u0000in plastic scenarios.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142248999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lars de Jong, Paula Clasen, Michael Müller, Ulrich Römer
In engineering, simulations play a vital role in predicting the behavior of a nonlinear dynamical system. In order to enhance the reliability of predictions, it is essential to incorporate the inherent uncertainties that are present in all real-world systems. Consequently, stochastic predictions are of significant importance, particularly during design or reliability analysis. In this work, we concentrate on the stochastic prediction of limit cycle oscillations, which typically occur in nonlinear dynamical systems and are of great technical importance. To address uncertainties in the limit cycle oscillations, we rely on the recently proposed Fourier generalized Polynomial Chaos expansion (FgPC), which combines Fourier analysis with spectral stochastic methods. In this paper, we demonstrate that valuable insights into the dynamics and their variability can be gained with a FgPC analysis, considering different benchmarks. These are the well-known forced Duffing oscillator and a more complex model from cell biology in which highly non-linear electrophysiological processes are closely linked to diffusive processes. With our spectral method, we are able to predict complicated marginal distributions of the limit cycle oscillations and, additionally, for self-excited systems, the uncertainty in the base frequency. Finally we study the sparsity of the FgPC coefficients as a basis for adaptive approximation.
{"title":"Uncertainty Analysis of Limit Cycle Oscillations in Nonlinear Dynamical Systems with the Fourier Generalized Polynomial Chaos Expansion","authors":"Lars de Jong, Paula Clasen, Michael Müller, Ulrich Römer","doi":"arxiv-2409.11006","DOIUrl":"https://doi.org/arxiv-2409.11006","url":null,"abstract":"In engineering, simulations play a vital role in predicting the behavior of a\u0000nonlinear dynamical system. In order to enhance the reliability of predictions,\u0000it is essential to incorporate the inherent uncertainties that are present in\u0000all real-world systems. Consequently, stochastic predictions are of significant\u0000importance, particularly during design or reliability analysis. In this work,\u0000we concentrate on the stochastic prediction of limit cycle oscillations, which\u0000typically occur in nonlinear dynamical systems and are of great technical\u0000importance. To address uncertainties in the limit cycle oscillations, we rely\u0000on the recently proposed Fourier generalized Polynomial Chaos expansion (FgPC),\u0000which combines Fourier analysis with spectral stochastic methods. In this\u0000paper, we demonstrate that valuable insights into the dynamics and their\u0000variability can be gained with a FgPC analysis, considering different\u0000benchmarks. These are the well-known forced Duffing oscillator and a more\u0000complex model from cell biology in which highly non-linear electrophysiological\u0000processes are closely linked to diffusive processes. With our spectral method,\u0000we are able to predict complicated marginal distributions of the limit cycle\u0000oscillations and, additionally, for self-excited systems, the uncertainty in\u0000the base frequency. Finally we study the sparsity of the FgPC coefficients as a\u0000basis for adaptive approximation.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142268202","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Noah M. Francis, Ricardo A. Lebensohn, Fatemeh Pourahmadian, Rémi Dingreville
This work presents a micromechanical spectral formulation for obtaining the full-field and homogenized response of elastoplastic micropolar composites. A closed-form radial-return mapping is derived from thermodynamics-based micropolar elastoplastic constitutive equations to determine the increment of plastic strain necessary to return the generalized stress state to the yield surface, and the algorithm implementation is verified using the method of numerically manufactured solutions. Then, size-dependent material response and micro-plasticity are shown as features that may be efficiently simulated in this micropolar elastoplastic framework. The computational efficiency of the formulation enables the generation of large datasets in reasonable computing times.
{"title":"Micropolar elastoplasticity using a fast Fourier transform-based solver","authors":"Noah M. Francis, Ricardo A. Lebensohn, Fatemeh Pourahmadian, Rémi Dingreville","doi":"arxiv-2409.10774","DOIUrl":"https://doi.org/arxiv-2409.10774","url":null,"abstract":"This work presents a micromechanical spectral formulation for obtaining the\u0000full-field and homogenized response of elastoplastic micropolar composites. A\u0000closed-form radial-return mapping is derived from thermodynamics-based\u0000micropolar elastoplastic constitutive equations to determine the increment of\u0000plastic strain necessary to return the generalized stress state to the yield\u0000surface, and the algorithm implementation is verified using the method of\u0000numerically manufactured solutions. Then, size-dependent material response and\u0000micro-plasticity are shown as features that may be efficiently simulated in\u0000this micropolar elastoplastic framework. The computational efficiency of the\u0000formulation enables the generation of large datasets in reasonable computing\u0000times.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142249043","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fast, gradient-based structural optimization has long been limited to a highly restricted subset of problems -- namely, density-based compliance minimization -- for which gradients can be analytically derived. For other objective functions, constraints, and design parameterizations, computing gradients has remained inaccessible, requiring the use of derivative-free algorithms that scale poorly with problem size. This has restricted the applicability of optimization to abstracted and academic problems, and has limited the uptake of these potentially impactful methods in practice. In this paper, we bridge the gap between computational efficiency and the freedom of problem formulation through a differentiable analysis framework designed for general structural optimization. We achieve this through leveraging Automatic Differentiation (AD) to manage the complex computational graph of structural analysis programs, and implementing specific derivation rules for performance critical functions along this graph. This paper provides a complete overview of gradient computation for arbitrary structural design objectives, identifies the barriers to their practical use, and derives key intermediate derivative operations that resolves these bottlenecks. Our framework is then tested against a series of structural design problems of increasing complexity: two highly constrained minimum volume problem, a multi-stage shape and section design problem, and an embodied carbon minimization problem. We benchmark our framework against other common optimization approaches, and show that our method outperforms others in terms of speed, stability, and solution quality.
{"title":"A differentiable structural analysis framework for high-performance design optimization","authors":"Keith J. Lee, Yijiang Huang, Caitlin T. Mueller","doi":"arxiv-2409.09247","DOIUrl":"https://doi.org/arxiv-2409.09247","url":null,"abstract":"Fast, gradient-based structural optimization has long been limited to a\u0000highly restricted subset of problems -- namely, density-based compliance\u0000minimization -- for which gradients can be analytically derived. For other\u0000objective functions, constraints, and design parameterizations, computing\u0000gradients has remained inaccessible, requiring the use of derivative-free\u0000algorithms that scale poorly with problem size. This has restricted the\u0000applicability of optimization to abstracted and academic problems, and has\u0000limited the uptake of these potentially impactful methods in practice. In this\u0000paper, we bridge the gap between computational efficiency and the freedom of\u0000problem formulation through a differentiable analysis framework designed for\u0000general structural optimization. We achieve this through leveraging Automatic\u0000Differentiation (AD) to manage the complex computational graph of structural\u0000analysis programs, and implementing specific derivation rules for performance\u0000critical functions along this graph. This paper provides a complete overview of\u0000gradient computation for arbitrary structural design objectives, identifies the\u0000barriers to their practical use, and derives key intermediate derivative\u0000operations that resolves these bottlenecks. Our framework is then tested\u0000against a series of structural design problems of increasing complexity: two\u0000highly constrained minimum volume problem, a multi-stage shape and section\u0000design problem, and an embodied carbon minimization problem. We benchmark our\u0000framework against other common optimization approaches, and show that our\u0000method outperforms others in terms of speed, stability, and solution quality.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142249045","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this work, we develop a new computational framework to solve sequential Bayesian experimental design (SBOED) problems constrained by large-scale partial differential equations with infinite-dimensional random parameters. We propose an adaptive terminal formulation of the optimality criteria for SBOED to achieve adaptive global optimality. We also establish an equivalent optimization formulation to achieve computational simplicity enabled by Laplace and low-rank approximations of the posterior. To accelerate the solution of the SBOED problem, we develop a derivative-informed latent attention neural operator (LANO), a new neural network surrogate model that leverages (1) derivative-informed dimension reduction for latent encoding, (2) an attention mechanism to capture the dynamics in the latent space, (3) an efficient training in the latent space augmented by projected Jacobian, which collectively lead to an efficient, accurate, and scalable surrogate in computing not only the parameter-to-observable (PtO) maps but also their Jacobians. We further develop the formulation for the computation of the MAP points, the eigenpairs, and the sampling from posterior by LANO in the reduced spaces and use these computations to solve the SBOED problem. We demonstrate the superior accuracy of LANO compared to two other neural architectures and the high accuracy of LANO compared to the finite element method (FEM) for the computation of MAP points in solving the SBOED problem with application to the experimental design of the time to take MRI images in monitoring tumor growth.
{"title":"Sequential infinite-dimensional Bayesian optimal experimental design with derivative-informed latent attention neural operator","authors":"Jinwoo Go, Peng Chen","doi":"arxiv-2409.09141","DOIUrl":"https://doi.org/arxiv-2409.09141","url":null,"abstract":"In this work, we develop a new computational framework to solve sequential\u0000Bayesian experimental design (SBOED) problems constrained by large-scale\u0000partial differential equations with infinite-dimensional random parameters. We\u0000propose an adaptive terminal formulation of the optimality criteria for SBOED\u0000to achieve adaptive global optimality. We also establish an equivalent\u0000optimization formulation to achieve computational simplicity enabled by Laplace\u0000and low-rank approximations of the posterior. To accelerate the solution of the\u0000SBOED problem, we develop a derivative-informed latent attention neural\u0000operator (LANO), a new neural network surrogate model that leverages (1)\u0000derivative-informed dimension reduction for latent encoding, (2) an attention\u0000mechanism to capture the dynamics in the latent space, (3) an efficient\u0000training in the latent space augmented by projected Jacobian, which\u0000collectively lead to an efficient, accurate, and scalable surrogate in\u0000computing not only the parameter-to-observable (PtO) maps but also their\u0000Jacobians. We further develop the formulation for the computation of the MAP\u0000points, the eigenpairs, and the sampling from posterior by LANO in the reduced\u0000spaces and use these computations to solve the SBOED problem. We demonstrate\u0000the superior accuracy of LANO compared to two other neural architectures and\u0000the high accuracy of LANO compared to the finite element method (FEM) for the\u0000computation of MAP points in solving the SBOED problem with application to the\u0000experimental design of the time to take MRI images in monitoring tumor growth.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142249046","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Muscles can store a large amount of genetic information, and in order to transform humans into computers, we need to start by increasing muscle tension. When people with cancer go on happy trips, some cancers often heal without treatment; Rhinitis can cause blockage of the nostrils, but after running, the nostrils naturally ventilate. Both are related to exercise, and the mystery behind them can treat both conditions. Cancer belongs to systemic diseases, and the eradication method for systemic diseases should start from the entire body system, treat the symptoms and prevent recurrence. This article uses special exercise methods and detailed methods to treat diseases, and finds that treating diseases from the perspective of the human system is indeed effective. This article adopts a comparative experimental method to compare the changes in the body before and after. Through this article, it is concluded that exercise and certain methods can cure mild rhinitis and promote rapid ventilation; Explaining from the perspective of muscle pulling force that older individuals are more prone to developing cellular variant cancer; Enhancing muscle tension in the human body can promote the cure of some cancers
{"title":"The mutual pulling force of human muscle fibers can treat mild cancer and rhinitis","authors":"Hongfa Zi, Ding Hua, Zhen Liu","doi":"arxiv-2409.08136","DOIUrl":"https://doi.org/arxiv-2409.08136","url":null,"abstract":"Muscles can store a large amount of genetic information, and in order to\u0000transform humans into computers, we need to start by increasing muscle tension.\u0000When people with cancer go on happy trips, some cancers often heal without\u0000treatment; Rhinitis can cause blockage of the nostrils, but after running, the\u0000nostrils naturally ventilate. Both are related to exercise, and the mystery\u0000behind them can treat both conditions. Cancer belongs to systemic diseases, and\u0000the eradication method for systemic diseases should start from the entire body\u0000system, treat the symptoms and prevent recurrence. This article uses special\u0000exercise methods and detailed methods to treat diseases, and finds that\u0000treating diseases from the perspective of the human system is indeed effective.\u0000This article adopts a comparative experimental method to compare the changes in\u0000the body before and after. Through this article, it is concluded that exercise\u0000and certain methods can cure mild rhinitis and promote rapid ventilation;\u0000Explaining from the perspective of muscle pulling force that older individuals\u0000are more prone to developing cellular variant cancer; Enhancing muscle tension\u0000in the human body can promote the cure of some cancers","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211142","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Efficiently retrieving an enormous chemical library to design targeted molecules is crucial for accelerating drug discovery, organic chemistry, and optoelectronic materials. Despite the emergence of generative models to produce novel drug-like molecules, in a more realistic scenario, the complexity of functional groups (e.g., pyrene, acenaphthylene, and bridged-ring systems) and extensive molecular scaffolds remain challenging obstacles for the generation of complex organics. Traditionally, the former demands an extra learning process, e.g., molecular pre-training, and the latter requires expensive computational resources. To address these challenges, we propose OrgMol-Design, a multi-granularity framework for efficiently designing complex organics. Our OrgMol-Design is composed of a score-based generative model via fragment prior for diverse coarse-grained scaffold generation and a chemical-rule-aware scoring model for fine-grained molecular structure design, circumventing the difficulty of intricate substructure learning without losing connection details among fragments. Our approach achieves state-of-the-art performance in four real-world and more challenging benchmarks covering broader scientific domains, outperforming advanced molecule generative models. Additionally, it delivers a substantial speedup and graphics memory reduction compared to diffusion-based graph models. Our results also demonstrate the importance of leveraging fragment prior for a generalized molecule inverse design model.
{"title":"Multi-granularity Score-based Generative Framework Enables Efficient Inverse Design of Complex Organics","authors":"Zijun Chen, Yu Wang, Liuzhenghao Lv, Hao Li, Zongying Lin, Li Yuan, Yonghong Tian","doi":"arxiv-2409.07912","DOIUrl":"https://doi.org/arxiv-2409.07912","url":null,"abstract":"Efficiently retrieving an enormous chemical library to design targeted\u0000molecules is crucial for accelerating drug discovery, organic chemistry, and\u0000optoelectronic materials. Despite the emergence of generative models to produce\u0000novel drug-like molecules, in a more realistic scenario, the complexity of\u0000functional groups (e.g., pyrene, acenaphthylene, and bridged-ring systems) and\u0000extensive molecular scaffolds remain challenging obstacles for the generation\u0000of complex organics. Traditionally, the former demands an extra learning\u0000process, e.g., molecular pre-training, and the latter requires expensive\u0000computational resources. To address these challenges, we propose OrgMol-Design,\u0000a multi-granularity framework for efficiently designing complex organics. Our\u0000OrgMol-Design is composed of a score-based generative model via fragment prior\u0000for diverse coarse-grained scaffold generation and a chemical-rule-aware\u0000scoring model for fine-grained molecular structure design, circumventing the\u0000difficulty of intricate substructure learning without losing connection details\u0000among fragments. Our approach achieves state-of-the-art performance in four\u0000real-world and more challenging benchmarks covering broader scientific domains,\u0000outperforming advanced molecule generative models. Additionally, it delivers a\u0000substantial speedup and graphics memory reduction compared to diffusion-based\u0000graph models. Our results also demonstrate the importance of leveraging\u0000fragment prior for a generalized molecule inverse design model.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Precise and timely forecasting of blood glucose levels is essential for effective diabetes management. While extensive research has been conducted on Type 1 diabetes mellitus, Type 2 diabetes mellitus (T2DM) presents unique challenges due to its heterogeneity, underscoring the need for specialized blood glucose forecasting systems. This study introduces a novel blood glucose forecasting system, applied to a dataset of 100 patients from the ShanghaiT2DM study. Our study uniquely integrates knowledge-driven and data-driven approaches, leveraging expert knowledge to validate and interpret the relationships among diabetes-related variables and deploying the data-driven approach to provide accurate forecast blood glucose levels. The Bayesian network approach facilitates the analysis of dependencies among various diabetes-related variables, thus enabling the inference of continuous glucose monitoring (CGM) trajectories in similar individuals with T2DM. By incorporating past CGM data including inference CGM trajectories, dietary records, and individual-specific information, the Bayesian structural time series (BSTS) model effectively forecasts glucose levels across time intervals ranging from 15 to 60 minutes. Forecast results show a mean absolute error of 6.41 mg/dL, a root mean square error of 8.29 mg/dL, and a mean absolute percentage error of 5.28%, for a 15-minute prediction horizon. This study makes the first application of the ShanghaiT2DM dataset for glucose level forecasting, considering the influences of diabetes-related variables. Its findings establish a foundational framework for developing personalized diabetes management strategies, potentially enhancing diabetes care through more accurate and timely interventions.
{"title":"Integrating Bayesian Approaches and Expert Knowledge for Forecasting Continuous Glucose Monitoring Values in Type 2 Diabetes Mellitus","authors":"Yuyang Sun, Panagiotis Kosmas","doi":"arxiv-2409.07315","DOIUrl":"https://doi.org/arxiv-2409.07315","url":null,"abstract":"Precise and timely forecasting of blood glucose levels is essential for\u0000effective diabetes management. While extensive research has been conducted on\u0000Type 1 diabetes mellitus, Type 2 diabetes mellitus (T2DM) presents unique\u0000challenges due to its heterogeneity, underscoring the need for specialized\u0000blood glucose forecasting systems. This study introduces a novel blood glucose\u0000forecasting system, applied to a dataset of 100 patients from the ShanghaiT2DM\u0000study. Our study uniquely integrates knowledge-driven and data-driven\u0000approaches, leveraging expert knowledge to validate and interpret the\u0000relationships among diabetes-related variables and deploying the data-driven\u0000approach to provide accurate forecast blood glucose levels. The Bayesian\u0000network approach facilitates the analysis of dependencies among various\u0000diabetes-related variables, thus enabling the inference of continuous glucose\u0000monitoring (CGM) trajectories in similar individuals with T2DM. By\u0000incorporating past CGM data including inference CGM trajectories, dietary\u0000records, and individual-specific information, the Bayesian structural time\u0000series (BSTS) model effectively forecasts glucose levels across time intervals\u0000ranging from 15 to 60 minutes. Forecast results show a mean absolute error of\u00006.41 mg/dL, a root mean square error of 8.29 mg/dL, and a mean absolute\u0000percentage error of 5.28%, for a 15-minute prediction horizon. This study makes\u0000the first application of the ShanghaiT2DM dataset for glucose level\u0000forecasting, considering the influences of diabetes-related variables. Its\u0000findings establish a foundational framework for developing personalized\u0000diabetes management strategies, potentially enhancing diabetes care through\u0000more accurate and timely interventions.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211144","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Microfluidic devices offer numerous advantages in medical applications, including the capture of single cells in microwell-based platforms for genomic analysis. As the cost of sequencing decreases, the demand for high-throughput single-cell analysis devices increases, leading to more microwells in a single device. However, their small size and large quantity increase the quality control (QC) effort. Currently, QC steps are still performed manually in some devices, requiring intensive training and time and causing inconsistency between different operators. A way to overcome this issue is to through automated defect detection. Computer vision can quickly analyze a large number of images in a short time and can be applied in defect detection. Automated defect detection can replace manual inspection, potentially decreasing variations in QC results. We report a machine learning (ML) algorithm that applies a convolution neural network (CNN) model with 9 layers and 64 units, incorporating dropouts and regularizations. This algorithm can analyze a large number of microwells produced by injection molding, significantly increasing the number of images analyzed compared to manual operator, improving QC, and ensuring the delivery of high-quality products to customers.
{"title":"Machine Learning Approaches for Defect Detection in a Microwell-based Medical Device","authors":"Xueying Zhao, Yan Chen, Yuefu Jiang, Amie Radenbaugh, Jamie Moskwa, Devon Jensen","doi":"arxiv-2409.07551","DOIUrl":"https://doi.org/arxiv-2409.07551","url":null,"abstract":"Microfluidic devices offer numerous advantages in medical applications,\u0000including the capture of single cells in microwell-based platforms for genomic\u0000analysis. As the cost of sequencing decreases, the demand for high-throughput\u0000single-cell analysis devices increases, leading to more microwells in a single\u0000device. However, their small size and large quantity increase the quality\u0000control (QC) effort. Currently, QC steps are still performed manually in some\u0000devices, requiring intensive training and time and causing inconsistency\u0000between different operators. A way to overcome this issue is to through\u0000automated defect detection. Computer vision can quickly analyze a large number\u0000of images in a short time and can be applied in defect detection. Automated\u0000defect detection can replace manual inspection, potentially decreasing\u0000variations in QC results. We report a machine learning (ML) algorithm that\u0000applies a convolution neural network (CNN) model with 9 layers and 64 units,\u0000incorporating dropouts and regularizations. This algorithm can analyze a large\u0000number of microwells produced by injection molding, significantly increasing\u0000the number of images analyzed compared to manual operator, improving QC, and\u0000ensuring the delivery of high-quality products to customers.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227894","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：