Min Woo Cho, Seok Hyeon Hwang, Jun-Young Jang, Jin Yeong Song, Sun-kwang Hwang, Kyoung Je Cha, Dong Yong Park, Kyungjun Song, Sang Min Park
Ventilated acoustic resonator(VAR), a type of acoustic metamaterial, emerge as an alternative for sound attenuation in environments that require ventilation, owing to its excellent low-frequency attenuation performance and flexible shape adaptability. However, due to the non-linear acoustic responses of VARs, the VAR designs are generally obtained within a limited parametrized design space, and the design relies on the iteration of the numerical simulation which consumes a considerable amount of computational time and resources. This paper proposes an acoustic response-encoded variational autoencoder (AR-VAE), a novel variational autoencoder-based generative design model for the efficient and accurate inverse design of VAR even with non-parametrized designs. The AR-VAE matches the high-dimensional acoustic response with the VAR cross-section image in the dimension-reduced latent space, which enables the AR-VAE to generate various non-parametrized VAR cross-section images with the target acoustic response. AR-VAE generates non-parameterized VARs from target acoustic responses, which show a 25-fold reduction in mean squared error compared to conventional deep learning-based parameter searching methods while exhibiting lower average mean squared error and peak frequency variance. By combining the inverse-designed VARs by AR-VAE, multi-cavity VAR was devised for broadband and multitarget peak frequency attenuation. The proposed design method presents a new approach for structural inverse-design with a high-dimensional non-linear physical response.
通风声共振（VAR）是一种声学超材料，由于其出色的低频衰减性能和灵活的形状适应性，在需要通风的环境中成为一种声音衰减的替代方法。然而，由于 VAR 的非线性声学响应，VAR 的设计通常只能在有限的参数化设计空间内获得，而且设计依赖于数值模拟的迭代，这将消耗大量的计算时间和资源。本文提出了一种声学响应编码变异自动编码器（AR-VAE），这是一种基于变异自动编码器的新型生成式设计模型，即使在非参数化设计的情况下也能高效、准确地进行 VAR 反设计。AR-VAE 将高维声学响应与降维潜在空间中的 VAR 横截面图像相匹配，从而使 AR-VAE 能够生成具有目标声学响应的各种非参数化 VAR 横截面图像。与基于深度学习的传统参数搜索方法相比，AR-VAE 从目标声学响应生成的非参数化 VAR 的均方误差降低了 25 倍，同时平均均方误差和峰值频率方差也更低。通过 AR-VAE 将逆向设计的 VAR 组合在一起，设计出了多腔 VAR，用于宽带和多目标峰值频率衰减。所提出的设计方法为具有高维非线性物理响应的结构逆设计提供了一种新方法。
{"title":"Inverse design of Non-parameterized Ventilated Acoustic Resonator via Variational Autoencoder with Acoustic Response-encoded Latent Space","authors":"Min Woo Cho, Seok Hyeon Hwang, Jun-Young Jang, Jin Yeong Song, Sun-kwang Hwang, Kyoung Je Cha, Dong Yong Park, Kyungjun Song, Sang Min Park","doi":"arxiv-2408.05917","DOIUrl":"https://doi.org/arxiv-2408.05917","url":null,"abstract":"Ventilated acoustic resonator(VAR), a type of acoustic metamaterial, emerge\u0000as an alternative for sound attenuation in environments that require\u0000ventilation, owing to its excellent low-frequency attenuation performance and\u0000flexible shape adaptability. However, due to the non-linear acoustic responses\u0000of VARs, the VAR designs are generally obtained within a limited parametrized\u0000design space, and the design relies on the iteration of the numerical\u0000simulation which consumes a considerable amount of computational time and\u0000resources. This paper proposes an acoustic response-encoded variational\u0000autoencoder (AR-VAE), a novel variational autoencoder-based generative design\u0000model for the efficient and accurate inverse design of VAR even with\u0000non-parametrized designs. The AR-VAE matches the high-dimensional acoustic\u0000response with the VAR cross-section image in the dimension-reduced latent\u0000space, which enables the AR-VAE to generate various non-parametrized VAR\u0000cross-section images with the target acoustic response. AR-VAE generates\u0000non-parameterized VARs from target acoustic responses, which show a 25-fold\u0000reduction in mean squared error compared to conventional deep learning-based\u0000parameter searching methods while exhibiting lower average mean squared error\u0000and peak frequency variance. By combining the inverse-designed VARs by AR-VAE,\u0000multi-cavity VAR was devised for broadband and multitarget peak frequency\u0000attenuation. The proposed design method presents a new approach for structural\u0000inverse-design with a high-dimensional non-linear physical response.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227901","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}
We introduce HyperCAN, a machine learning framework that utilizes hypernetworks to construct adaptable constitutive artificial neural networks for a wide range of beam-based metamaterials exhibiting diverse mechanical behavior under finite deformations. HyperCAN integrates an input convex network that models the nonlinear stress-strain map of a truss lattice, while ensuring adherence to fundamental mechanics principles, along with a hypernetwork that dynamically adjusts the parameters of the convex network as a function of the lattice topology and geometry. This unified framework demonstrates robust generalization in predicting the mechanical behavior of previously unseen metamaterial designs and loading scenarios well beyond the training domain. We show how HyperCAN can be integrated into multiscale simulations to accurately capture the highly nonlinear responses of large-scale truss metamaterials, closely matching fully resolved simulations while significantly reducing computational costs. This offers new efficient opportunities for the multiscale design and optimization of truss metamaterials.
{"title":"HyperCAN: Hypernetwork-Driven Deep Parameterized Constitutive Models for Metamaterials","authors":"Li Zheng, Dennis M. Kochmann, Siddhant Kumar","doi":"arxiv-2408.06017","DOIUrl":"https://doi.org/arxiv-2408.06017","url":null,"abstract":"We introduce HyperCAN, a machine learning framework that utilizes\u0000hypernetworks to construct adaptable constitutive artificial neural networks\u0000for a wide range of beam-based metamaterials exhibiting diverse mechanical\u0000behavior under finite deformations. HyperCAN integrates an input convex network\u0000that models the nonlinear stress-strain map of a truss lattice, while ensuring\u0000adherence to fundamental mechanics principles, along with a hypernetwork that\u0000dynamically adjusts the parameters of the convex network as a function of the\u0000lattice topology and geometry. This unified framework demonstrates robust\u0000generalization in predicting the mechanical behavior of previously unseen\u0000metamaterial designs and loading scenarios well beyond the training domain. We\u0000show how HyperCAN can be integrated into multiscale simulations to accurately\u0000capture the highly nonlinear responses of large-scale truss metamaterials,\u0000closely matching fully resolved simulations while significantly reducing\u0000computational costs. This offers new efficient opportunities for the multiscale\u0000design and optimization of truss metamaterials.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211345","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}
Tim van der Velden, Stefanie Reese, Hagen Holthusen, Tim Brepols
This paper establishes a universal framework for the nonlocal modeling of anisotropic damage at finite strains. By the combination of two recent works, the new framework allows for the flexible incorporation of different established hyperelastic finite strain material formulations into anisotropic damage whilst ensuring mesh-independent results by employing a generic set of micromorphic gradient-extensions. First, the anisotropic damage model, generally satisfying the damage growth criterion, is investigated for the specific choice of a Neo-Hookean material on a single element. Next, the model is applied with different gradient-extensions in structural simulations of an asymmetrically notched specimen to identify an efficient choice in the form of a volumetric-deviatoric regularization. Thereafter, the universal framework, which is without loss of generality here specified for a Neo-Hookean material with a volumetric-deviatoric gradient-extension, successfully serves for the complex simulation of a pressure loaded rotor blade. After acceptance of the manuscript, we make the codes of the material subroutines accessible to the public at https://doi.org/10.5281/zenodo.11171630.
{"title":"An anisotropic, brittle damage model for finite strains with a generic damage tensor regularization","authors":"Tim van der Velden, Stefanie Reese, Hagen Holthusen, Tim Brepols","doi":"arxiv-2408.06140","DOIUrl":"https://doi.org/arxiv-2408.06140","url":null,"abstract":"This paper establishes a universal framework for the nonlocal modeling of\u0000anisotropic damage at finite strains. By the combination of two recent works,\u0000the new framework allows for the flexible incorporation of different\u0000established hyperelastic finite strain material formulations into anisotropic\u0000damage whilst ensuring mesh-independent results by employing a generic set of\u0000micromorphic gradient-extensions. First, the anisotropic damage model,\u0000generally satisfying the damage growth criterion, is investigated for the\u0000specific choice of a Neo-Hookean material on a single element. Next, the model\u0000is applied with different gradient-extensions in structural simulations of an\u0000asymmetrically notched specimen to identify an efficient choice in the form of\u0000a volumetric-deviatoric regularization. Thereafter, the universal framework,\u0000which is without loss of generality here specified for a Neo-Hookean material\u0000with a volumetric-deviatoric gradient-extension, successfully serves for the\u0000complex simulation of a pressure loaded rotor blade. After acceptance of the manuscript, we make the codes of the material\u0000subroutines accessible to the public at\u0000https://doi.org/10.5281/zenodo.11171630.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211347","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}
Xirui Tang, Zeyu Wang, Xiaowei Cai, Honghua Su, Changsong Wei
The rapid development of the mobile Internet and the Internet of Things is leading to a diversification of user devices and the emergence of new mobile applications on a regular basis. Such applications include those that are computationally intensive, such as pattern recognition, interactive gaming, virtual reality, and augmented reality. However, the computing and energy resources available on the user's equipment are limited, which presents a challenge in effectively supporting such demanding applications. In this work, we propose a heterogeneous computing resource allocation model based on a data-driven approach. The model first collects and analyzes historical workload data at scale, extracts key features, and builds a detailed data set. Then, a data-driven deep neural network is used to predict future resource requirements. Based on the prediction results, the model adopts a dynamic adjustment and optimization resource allocation strategy. This strategy not only fully considers the characteristics of different computing resources, but also accurately matches the requirements of various tasks, and realizes dynamic and flexible resource allocation, thereby greatly improving the overall performance and resource utilization of the system. Experimental results show that the proposed method is significantly better than the traditional resource allocation method in a variety of scenarios, demonstrating its excellent accuracy and adaptability.
{"title":"Research on Heterogeneous Computation Resource Allocation based on Data-driven Method","authors":"Xirui Tang, Zeyu Wang, Xiaowei Cai, Honghua Su, Changsong Wei","doi":"arxiv-2408.05671","DOIUrl":"https://doi.org/arxiv-2408.05671","url":null,"abstract":"The rapid development of the mobile Internet and the Internet of Things is\u0000leading to a diversification of user devices and the emergence of new mobile\u0000applications on a regular basis. Such applications include those that are\u0000computationally intensive, such as pattern recognition, interactive gaming,\u0000virtual reality, and augmented reality. However, the computing and energy\u0000resources available on the user's equipment are limited, which presents a\u0000challenge in effectively supporting such demanding applications. In this work,\u0000we propose a heterogeneous computing resource allocation model based on a\u0000data-driven approach. The model first collects and analyzes historical workload\u0000data at scale, extracts key features, and builds a detailed data set. Then, a\u0000data-driven deep neural network is used to predict future resource\u0000requirements. Based on the prediction results, the model adopts a dynamic\u0000adjustment and optimization resource allocation strategy. This strategy not\u0000only fully considers the characteristics of different computing resources, but\u0000also accurately matches the requirements of various tasks, and realizes dynamic\u0000and flexible resource allocation, thereby greatly improving the overall\u0000performance and resource utilization of the system. Experimental results show\u0000that the proposed method is significantly better than the traditional resource\u0000allocation method in a variety of scenarios, demonstrating its excellent\u0000accuracy and adaptability.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211346","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}
Sebastian Rodriguez, Angelo Pasquale, Jad Mounayer, Diego Canales, Marianne Beringhier, Chady Ghnatios, Amine Ammar, Francisco Chinesta
The simulation of viscoelastic time-evolution problems described by a large number of internal variables and with a large spectrum of relaxation times requires high computational resources for their resolution. Furthermore, the internal variables evolution is described by a set of linear differential equations which involves many time scales. In this context, the use of a space-time PGD approximation is proposed here to boost their resolution, where the temporal functions are constructed following a multi-scale strategy along with the Partition of Unity method, in order to catch each dynamic efficiently. The feasibility and the robustness of the method are discussed in the case of a polymer in a non-equilibrium state under cyclic loading.
{"title":"A reduced simulation applied to viscoelastic fatigue of polymers using a time multi-scale approach based on Partition of Unity method","authors":"Sebastian Rodriguez, Angelo Pasquale, Jad Mounayer, Diego Canales, Marianne Beringhier, Chady Ghnatios, Amine Ammar, Francisco Chinesta","doi":"arxiv-2408.05143","DOIUrl":"https://doi.org/arxiv-2408.05143","url":null,"abstract":"The simulation of viscoelastic time-evolution problems described by a large\u0000number of internal variables and with a large spectrum of relaxation times\u0000requires high computational resources for their resolution. Furthermore, the\u0000internal variables evolution is described by a set of linear differential\u0000equations which involves many time scales. In this context, the use of a\u0000space-time PGD approximation is proposed here to boost their resolution, where\u0000the temporal functions are constructed following a multi-scale strategy along\u0000with the Partition of Unity method, in order to catch each dynamic efficiently.\u0000The feasibility and the robustness of the method are discussed in the case of a\u0000polymer in a non-equilibrium state under cyclic loading.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141936369","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}
Sohini Roychowdhury, Marko Krema, Brian Moore, Xingjian Lai, Dike Effedua, Bharat Jethwani
Financial report generation using general purpose large language models pose two major challenges, including the lack of compound sentences and hallucinations. Advanced prompt engineering and retrieval augmented generation (RAG) techniques are incapable of curing the writing style discrepancies. In this work we propose a novel two-stage fine-tuning process wherein public domain financial reports are processed into prompt-completions and augmented using simple LLM prompts to then enable sectional financial report generation using minimal instructions and tabular data inputs. Our proposed fine-tuning framework results doubles the number of correct questions answers and reduces hallucinations by over 50%. Additionally, the two-stage fine tuned models have lower perplexity, improved ROUGE, TER and BLEU scores, higher creativity and knowledge density with lower uncertainty and cross entropy.
{"title":"FiST-Financial Style Transfer with Hallucination and Creativity Control Framework","authors":"Sohini Roychowdhury, Marko Krema, Brian Moore, Xingjian Lai, Dike Effedua, Bharat Jethwani","doi":"arxiv-2408.05365","DOIUrl":"https://doi.org/arxiv-2408.05365","url":null,"abstract":"Financial report generation using general purpose large language models pose\u0000two major challenges, including the lack of compound sentences and\u0000hallucinations. Advanced prompt engineering and retrieval augmented generation\u0000(RAG) techniques are incapable of curing the writing style discrepancies. In\u0000this work we propose a novel two-stage fine-tuning process wherein public\u0000domain financial reports are processed into prompt-completions and augmented\u0000using simple LLM prompts to then enable sectional financial report generation\u0000using minimal instructions and tabular data inputs. Our proposed fine-tuning\u0000framework results doubles the number of correct questions answers and reduces\u0000hallucinations by over 50%. Additionally, the two-stage fine tuned models have\u0000lower perplexity, improved ROUGE, TER and BLEU scores, higher creativity and\u0000knowledge density with lower uncertainty and cross entropy.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211350","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}
Various machine learning (ML)-based in-situ monitoring systems have been developed to detect laser additive manufacturing (LAM) process anomalies and defects. Multimodal fusion can improve in-situ monitoring performance by acquiring and integrating data from multiple modalities, including visual and audio data. However, multimodal fusion employs multiple sensors of different types, which leads to higher hardware, computational, and operational costs. This paper proposes a cross-modality knowledge transfer (CMKT) methodology that transfers knowledge from a source to a target modality for LAM in-situ monitoring. CMKT enhances the usefulness of the features extracted from the target modality during the training phase and removes the sensors of the source modality during the prediction phase. This paper proposes three CMKT methods: semantic alignment, fully supervised mapping, and semi-supervised mapping. Semantic alignment establishes a shared encoded space between modalities to facilitate knowledge transfer. It utilizes a semantic alignment loss to align the distributions of the same classes (e.g., visual defective and audio defective classes) and a separation loss to separate the distributions of different classes (e.g., visual defective and audio defect-free classes). The two mapping methods transfer knowledge by deriving the features of one modality from the other modality using fully supervised and semi-supervised learning. The proposed CMKT methods were implemented and compared with multimodal audio-visual fusion in an LAM in-situ anomaly detection case study. The semantic alignment method achieves a 98.4% accuracy while removing the audio modality during the prediction phase, which is comparable to the accuracy of multimodal fusion (98.2%).
{"title":"Audio-visual cross-modality knowledge transfer for machine learning-based in-situ monitoring in laser additive manufacturing","authors":"Jiarui Xie, Mutahar Safdar, Lequn Chen, Seung Ki Moon, Yaoyao Fiona Zhao","doi":"arxiv-2408.05307","DOIUrl":"https://doi.org/arxiv-2408.05307","url":null,"abstract":"Various machine learning (ML)-based in-situ monitoring systems have been\u0000developed to detect laser additive manufacturing (LAM) process anomalies and\u0000defects. Multimodal fusion can improve in-situ monitoring performance by\u0000acquiring and integrating data from multiple modalities, including visual and\u0000audio data. However, multimodal fusion employs multiple sensors of different\u0000types, which leads to higher hardware, computational, and operational costs.\u0000This paper proposes a cross-modality knowledge transfer (CMKT) methodology that\u0000transfers knowledge from a source to a target modality for LAM in-situ\u0000monitoring. CMKT enhances the usefulness of the features extracted from the\u0000target modality during the training phase and removes the sensors of the source\u0000modality during the prediction phase. This paper proposes three CMKT methods:\u0000semantic alignment, fully supervised mapping, and semi-supervised mapping.\u0000Semantic alignment establishes a shared encoded space between modalities to\u0000facilitate knowledge transfer. It utilizes a semantic alignment loss to align\u0000the distributions of the same classes (e.g., visual defective and audio\u0000defective classes) and a separation loss to separate the distributions of\u0000different classes (e.g., visual defective and audio defect-free classes). The\u0000two mapping methods transfer knowledge by deriving the features of one modality\u0000from the other modality using fully supervised and semi-supervised learning.\u0000The proposed CMKT methods were implemented and compared with multimodal\u0000audio-visual fusion in an LAM in-situ anomaly detection case study. The\u0000semantic alignment method achieves a 98.4% accuracy while removing the audio\u0000modality during the prediction phase, which is comparable to the accuracy of\u0000multimodal fusion (98.2%).","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211348","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}
Björn Lütjens, Raffaele Ferrari, Duncan Watson-Parris, Noelle Selin
Full-complexity Earth system models (ESMs) are computationally very expensive, limiting their use in exploring the climate outcomes of multiple emission pathways. More efficient emulators that approximate ESMs can directly map emissions onto climate outcomes, and benchmarks are being used to evaluate their accuracy on standardized tasks and datasets. We investigate a popular benchmark in data-driven climate emulation, ClimateBench, on which deep learning-based emulators are currently achieving the best performance. We implement a linear regression-based emulator, akin to pattern scaling, and find that it outperforms the incumbent 100M-parameter deep learning foundation model, ClimaX, on 3 out of 4 regionally-resolved surface-level climate variables. While emulating surface temperature is expected to be predominantly linear, this result is surprising for emulating precipitation. We identify that this outcome is a result of high levels of internal variability in the benchmark targets. To address internal variability, we update the benchmark targets with ensemble averages from the MPI-ESM1.2-LR model that contain 50 instead of 3 climate simulations per emission pathway. Using the new targets, we show that linear pattern scaling continues to be more accurate on temperature, but can be outperformed by a deep learning-based model for emulating precipitation. We publish our code, data, and an interactive tutorial at github.com/blutjens/climate-emulator.
{"title":"The impact of internal variability on benchmarking deep learning climate emulators","authors":"Björn Lütjens, Raffaele Ferrari, Duncan Watson-Parris, Noelle Selin","doi":"arxiv-2408.05288","DOIUrl":"https://doi.org/arxiv-2408.05288","url":null,"abstract":"Full-complexity Earth system models (ESMs) are computationally very\u0000expensive, limiting their use in exploring the climate outcomes of multiple\u0000emission pathways. More efficient emulators that approximate ESMs can directly\u0000map emissions onto climate outcomes, and benchmarks are being used to evaluate\u0000their accuracy on standardized tasks and datasets. We investigate a popular\u0000benchmark in data-driven climate emulation, ClimateBench, on which deep\u0000learning-based emulators are currently achieving the best performance. We\u0000implement a linear regression-based emulator, akin to pattern scaling, and find\u0000that it outperforms the incumbent 100M-parameter deep learning foundation\u0000model, ClimaX, on 3 out of 4 regionally-resolved surface-level climate\u0000variables. While emulating surface temperature is expected to be predominantly\u0000linear, this result is surprising for emulating precipitation. We identify that\u0000this outcome is a result of high levels of internal variability in the\u0000benchmark targets. To address internal variability, we update the benchmark\u0000targets with ensemble averages from the MPI-ESM1.2-LR model that contain 50\u0000instead of 3 climate simulations per emission pathway. Using the new targets,\u0000we show that linear pattern scaling continues to be more accurate on\u0000temperature, but can be outperformed by a deep learning-based model for\u0000emulating precipitation. We publish our code, data, and an interactive tutorial\u0000at github.com/blutjens/climate-emulator.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211349","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}
Sebastian Rodriguez, Pierre-Etienne Charbonnel, Pierre Ladevèze, David Néron
This paper presents a first implementation of the LArge Time INcrement (LATIN) method along with the model reduction technique called Proper Generalized Decomposition (PGD) for solving nonlinear low-frequency dynamics problems when dealing with a quasi-brittle isotropic damage constitutive relations. The present paper uses the Time-Discontinuous Galerkin Method (TDGM) for computing the temporal contributions of the space-time separate-variables solution of the LATIN-PGD approach, which offers several advantages when considering a high number of DOFs in time. The efficiency of the method is tested for the case of a 3D bending beam, where results and benchmarks comparing LATIN-PGD to classical time-incremental Newmark/Quasi-Newton nonlinear solver are presented. This work represents a first step towards taking into account uncertainties and carrying out more complex parametric studies imposed by seismic risk assessment.
本文首次提出了LArge Time INcrement (LATIN)方法与称为ProperGeneralized Decomposition (PGD)的模型缩减技术，用于求解准脆性各向同性损伤构成参数时的非线性低频动力学问题。本文使用时间-非连续伽勒金方法（TDGM）计算 LATIN-PGD 方法的时空分离无变量求解的时间贡献。以三维弯曲梁为例，测试了该方法的效率，并给出了将 LATIN-PGD 与经典的时间递增纽马克/准牛顿非线性求解器进行比较的结果和基准。这项工作是考虑不确定性和开展地震风险评估所要求的更复杂参数研究的第一步。
{"title":"The LATIN-PGD methodology to nonlinear dynamics and quasi-brittle materials for future earthquake engineering applications","authors":"Sebastian Rodriguez, Pierre-Etienne Charbonnel, Pierre Ladevèze, David Néron","doi":"arxiv-2408.05108","DOIUrl":"https://doi.org/arxiv-2408.05108","url":null,"abstract":"This paper presents a first implementation of the LArge Time INcrement\u0000(LATIN) method along with the model reduction technique called Proper\u0000Generalized Decomposition (PGD) for solving nonlinear low-frequency dynamics\u0000problems when dealing with a quasi-brittle isotropic damage constitutive\u0000relations. The present paper uses the Time-Discontinuous Galerkin Method (TDGM)\u0000for computing the temporal contributions of the space-time separate-variables\u0000solution of the LATIN-PGD approach, which offers several advantages when\u0000considering a high number of DOFs in time. The efficiency of the method is\u0000tested for the case of a 3D bending beam, where results and benchmarks\u0000comparing LATIN-PGD to classical time-incremental Newmark/Quasi-Newton\u0000nonlinear solver are presented. This work represents a first step towards\u0000taking into account uncertainties and carrying out more complex parametric\u0000studies imposed by seismic risk assessment.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141936368","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}
The burgeoning field of autonomous driving necessitates the seamless integration of autonomous vehicles (AVs) with human-driven vehicles, calling for more predictable AV behavior and enhanced interaction with human drivers. Human-like driving, particularly during lane-changing maneuvers on highways, is a critical area of research due to its significant impact on safety and traffic flow. Traditional rule-based decision-making approaches often fail to encapsulate the nuanced boundaries of human behavior in diverse driving scenarios, while crafting reward functions for learning-based methods introduces its own set of complexities. This study investigates the application of Reinforcement Learning from Human Feedback (RLHF) to emulate human-like lane-changing decisions in AVs. An initial RL policy is pre-trained to ensure safe lane changes. Subsequently, this policy is employed to gather data, which is then annotated by humans to train a reward model that discerns lane changes aligning with human preferences. This human-informed reward model supersedes the original, guiding the refinement of the policy to reflect human-like preferences. The effectiveness of RLHF in producing human-like lane changes is demonstrated through the development and evaluation of conservative and aggressive lane-changing models within obstacle-rich environments and mixed autonomy traffic scenarios. The experimental outcomes underscore the potential of RLHF to diversify lane-changing behaviors in AVs, suggesting its viability for enhancing the integration of AVs into the fabric of human-driven traffic.
{"title":"Reinforcement Learning from Human Feedback for Lane Changing of Autonomous Vehicles in Mixed Traffic","authors":"Yuting Wang, Lu Liu, Maonan Wang, Xi Xiong","doi":"arxiv-2408.04447","DOIUrl":"https://doi.org/arxiv-2408.04447","url":null,"abstract":"The burgeoning field of autonomous driving necessitates the seamless\u0000integration of autonomous vehicles (AVs) with human-driven vehicles, calling\u0000for more predictable AV behavior and enhanced interaction with human drivers.\u0000Human-like driving, particularly during lane-changing maneuvers on highways, is\u0000a critical area of research due to its significant impact on safety and traffic\u0000flow. Traditional rule-based decision-making approaches often fail to\u0000encapsulate the nuanced boundaries of human behavior in diverse driving\u0000scenarios, while crafting reward functions for learning-based methods\u0000introduces its own set of complexities. This study investigates the application\u0000of Reinforcement Learning from Human Feedback (RLHF) to emulate human-like\u0000lane-changing decisions in AVs. An initial RL policy is pre-trained to ensure\u0000safe lane changes. Subsequently, this policy is employed to gather data, which\u0000is then annotated by humans to train a reward model that discerns lane changes\u0000aligning with human preferences. This human-informed reward model supersedes\u0000the original, guiding the refinement of the policy to reflect human-like\u0000preferences. The effectiveness of RLHF in producing human-like lane changes is\u0000demonstrated through the development and evaluation of conservative and\u0000aggressive lane-changing models within obstacle-rich environments and mixed\u0000autonomy traffic scenarios. The experimental outcomes underscore the potential\u0000of RLHF to diversify lane-changing behaviors in AVs, suggesting its viability\u0000for enhancing the integration of AVs into the fabric of human-driven traffic.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141936374","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：