The thermal conductivity of covalent organic frameworks (COFs), an emerging class of nanoporous polymeric materials, is crucial for many applications, yet the link between their structure and thermal properties is not well understood. From a dataset of over 2,400 COFs, we find that conventional features like density, pore size, void fraction, and surface area do not reliably predict thermal conductivity. To overcome this, we train an attention-based machine learning model that accurately predicts thermal conductivities, even for structures outside the training set. We then use the attention mechanism to understand why the model works. Surprisingly, dangling molecular branches emerge as key predictors of thermal conductivity, alongside conventional geometric descriptors like density and pore size. Our findings show that COFs with dangling functional groups exhibit lower thermal transfer capabilities than otherwise. Molecular dynamics simulations confirm this, revealing significant mismatches in the vibrational density of states due to the presence of dangling branches.
{"title":"Deep learning reveals key predictors of thermal conductivity in covalent organic frameworks","authors":"Prakash Thakolkaran, Yiwen Zheng, Yaqi Guo, Aniruddh Vashisth, Siddhant Kumar","doi":"arxiv-2409.06457","DOIUrl":"https://doi.org/arxiv-2409.06457","url":null,"abstract":"The thermal conductivity of covalent organic frameworks (COFs), an emerging\u0000class of nanoporous polymeric materials, is crucial for many applications, yet\u0000the link between their structure and thermal properties is not well understood.\u0000From a dataset of over 2,400 COFs, we find that conventional features like\u0000density, pore size, void fraction, and surface area do not reliably predict\u0000thermal conductivity. To overcome this, we train an attention-based machine\u0000learning model that accurately predicts thermal conductivities, even for\u0000structures outside the training set. We then use the attention mechanism to\u0000understand why the model works. Surprisingly, dangling molecular branches\u0000emerge as key predictors of thermal conductivity, alongside conventional\u0000geometric descriptors like density and pore size. Our findings show that COFs\u0000with dangling functional groups exhibit lower thermal transfer capabilities\u0000than otherwise. Molecular dynamics simulations confirm this, revealing\u0000significant mismatches in the vibrational density of states due to the presence\u0000of dangling branches.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211147","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}
Sajjad Arzemanzadeh, Benjamin Zwick, Karol Miller, Tim Rosenow, Stuart I. Hodgetts, Adam Wittek
Computational biomechanics models of the brain have become an important tool for investigating the brain responses to mechanical loads. The geometry, loading conditions, and constitutive properties of such brain models are well-studied and generally accepted. However, there is a lack of experimental evidence to support models of the layers of tissues (brain-skull interface) connecting the brain with the skull which determine boundary conditions for the brain. We present a new protocol for determining the biomechanical properties of the brain-skull interface and present the preliminary results (for a small number of tissue samples extracted from sheep cadaver heads). The method consists of biomechanical experiments using brain tissue and brain-skull complex (consisting of the brain tissue, brain-skull interface, and skull bone) and comprehensive computer simulation of the experiments using the finite element (FE) method. Application of the FE simulations allowed us to abandon the traditionally used approaches that rely on analytical formulations that assume cuboidal (or cylindrical) sample geometry when determining the parameters that describe the biomechanical behaviour of the brain tissue and brain-skull interface. In the simulations, we used accurate 3D geometry of the samples obtained from magnetic resonance images (MRIs). Our results indicate that the behaviour of the brain-skull interface under compressive loading appreciably differs from that under tension. Rupture of the interface was clearly visible for tensile load while no obvious indication of mechanical failure was observed under compression. These results suggest that assuming a rigid connection or frictionless sliding contact between the brain tissue and skull bone, the approaches often used in computational biomechanics models of the brain, may not accurately represent the mechanical behaviour of the brain-skull interface.
{"title":"Towards Determining Mechanical Properties of Brain-Skull Interface Under Tension and Compression","authors":"Sajjad Arzemanzadeh, Benjamin Zwick, Karol Miller, Tim Rosenow, Stuart I. Hodgetts, Adam Wittek","doi":"arxiv-2409.05365","DOIUrl":"https://doi.org/arxiv-2409.05365","url":null,"abstract":"Computational biomechanics models of the brain have become an important tool\u0000for investigating the brain responses to mechanical loads. The geometry,\u0000loading conditions, and constitutive properties of such brain models are\u0000well-studied and generally accepted. However, there is a lack of experimental\u0000evidence to support models of the layers of tissues (brain-skull interface)\u0000connecting the brain with the skull which determine boundary conditions for the\u0000brain. We present a new protocol for determining the biomechanical properties\u0000of the brain-skull interface and present the preliminary results (for a small\u0000number of tissue samples extracted from sheep cadaver heads). The method\u0000consists of biomechanical experiments using brain tissue and brain-skull\u0000complex (consisting of the brain tissue, brain-skull interface, and skull bone)\u0000and comprehensive computer simulation of the experiments using the finite\u0000element (FE) method. Application of the FE simulations allowed us to abandon\u0000the traditionally used approaches that rely on analytical formulations that\u0000assume cuboidal (or cylindrical) sample geometry when determining the\u0000parameters that describe the biomechanical behaviour of the brain tissue and\u0000brain-skull interface. In the simulations, we used accurate 3D geometry of the\u0000samples obtained from magnetic resonance images (MRIs). Our results indicate\u0000that the behaviour of the brain-skull interface under compressive loading\u0000appreciably differs from that under tension. Rupture of the interface was\u0000clearly visible for tensile load while no obvious indication of mechanical\u0000failure was observed under compression. These results suggest that assuming a\u0000rigid connection or frictionless sliding contact between the brain tissue and\u0000skull bone, the approaches often used in computational biomechanics models of\u0000the brain, may not accurately represent the mechanical behaviour of the\u0000brain-skull interface.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211145","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}
Fu Lin, Jiasheng Shi, Shijie Luo, Qinpei Zhao, Weixiong Rao, Lei Chen
Traditional simulation of complex mechanical systems relies on numerical solvers of Partial Differential Equations (PDEs), e.g., using the Finite Element Method (FEM). The FEM solvers frequently suffer from intensive computation cost and high running time. Recent graph neural network (GNN)-based simulation models can improve running time meanwhile with acceptable accuracy. Unfortunately, they are hard to tailor GNNs for complex mechanical systems, including such disadvantages as ineffective representation and inefficient message propagation (MP). To tackle these issues, in this paper, with the proposed Up-sampling-only and Adaptive MP techniques, we develop a novel hierarchical Mesh Graph Network, namely UA-MGN, for efficient and effective mechanical simulation. Evaluation on two synthetic and one real datasets demonstrates the superiority of the UA-MGN. For example, on the Beam dataset, compared to the state-of-the-art MS-MGN, UA-MGN leads to 40.99% lower errors but using only 43.48% fewer network parameters and 4.49% fewer floating point operations (FLOPs).
{"title":"Up-sampling-only and Adaptive Mesh-based GNN for Simulating Physical Systems","authors":"Fu Lin, Jiasheng Shi, Shijie Luo, Qinpei Zhao, Weixiong Rao, Lei Chen","doi":"arxiv-2409.04740","DOIUrl":"https://doi.org/arxiv-2409.04740","url":null,"abstract":"Traditional simulation of complex mechanical systems relies on numerical\u0000solvers of Partial Differential Equations (PDEs), e.g., using the Finite\u0000Element Method (FEM). The FEM solvers frequently suffer from intensive\u0000computation cost and high running time. Recent graph neural network (GNN)-based\u0000simulation models can improve running time meanwhile with acceptable accuracy.\u0000Unfortunately, they are hard to tailor GNNs for complex mechanical systems,\u0000including such disadvantages as ineffective representation and inefficient\u0000message propagation (MP). To tackle these issues, in this paper, with the\u0000proposed Up-sampling-only and Adaptive MP techniques, we develop a novel\u0000hierarchical Mesh Graph Network, namely UA-MGN, for efficient and effective\u0000mechanical simulation. Evaluation on two synthetic and one real datasets\u0000demonstrates the superiority of the UA-MGN. For example, on the Beam dataset,\u0000compared to the state-of-the-art MS-MGN, UA-MGN leads to 40.99% lower errors\u0000but using only 43.48% fewer network parameters and 4.49% fewer floating point\u0000operations (FLOPs).","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211146","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}
Pasquale De Rosa, Pascal Felber, Valerio Schiavoni
This paper introduces CryptoAnalytics, a software toolkit for cryptocoins price forecasting with machine learning (ML) techniques. Cryptocoins are tradable digital assets exchanged for specific trading prices. While history has shown the extreme volatility of such trading prices, the ability to efficiently model and forecast the time series resulting from the exchange price volatility remains an open research challenge. Good results can been achieved with state-of-the-art ML techniques, including Gradient-Boosting Machines (GBMs) and Recurrent Neural Networks (RNNs). CryptoAnalytics is a software toolkit to easily train these models and make inference on up-to-date cryptocoin trading price data, with facilities to fetch datasets from one of the main leading aggregator websites, i.e., CoinMarketCap, train models and infer the future trends. This software is implemented in Python. It relies on PyTorch for the implementation of RNNs (LSTM and GRU), while for GBMs, it leverages on XgBoost, LightGBM and CatBoost.
{"title":"CryptoAnalytics: Cryptocoins Price Forecasting with Machine Learning Techniques","authors":"Pasquale De Rosa, Pascal Felber, Valerio Schiavoni","doi":"arxiv-2409.04106","DOIUrl":"https://doi.org/arxiv-2409.04106","url":null,"abstract":"This paper introduces CryptoAnalytics, a software toolkit for cryptocoins\u0000price forecasting with machine learning (ML) techniques. Cryptocoins are\u0000tradable digital assets exchanged for specific trading prices. While history\u0000has shown the extreme volatility of such trading prices, the ability to\u0000efficiently model and forecast the time series resulting from the exchange\u0000price volatility remains an open research challenge. Good results can been\u0000achieved with state-of-the-art ML techniques, including Gradient-Boosting\u0000Machines (GBMs) and Recurrent Neural Networks (RNNs). CryptoAnalytics is a\u0000software toolkit to easily train these models and make inference on up-to-date\u0000cryptocoin trading price data, with facilities to fetch datasets from one of\u0000the main leading aggregator websites, i.e., CoinMarketCap, train models and\u0000infer the future trends. This software is implemented in Python. It relies on\u0000PyTorch for the implementation of RNNs (LSTM and GRU), while for GBMs, it\u0000leverages on XgBoost, LightGBM and CatBoost.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227896","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}
Pasquale De Rosa, Pascal Felber, Valerio Schiavoni
Cryptocoins (i.e., Bitcoin, Ether, Litecoin) are tradable digital assets. Ownerships of cryptocoins are registered on distributed ledgers (i.e., blockchains). Secure encryption techniques guarantee the security of the transactions (transfers of coins among owners), registered into the ledger. Cryptocoins are exchanged for specific trading prices. The extreme volatility of such trading prices across all different sets of crypto-assets remains undisputed. However, the relations between the trading prices across different cryptocoins remains largely unexplored. Major coin exchanges indicate trend correlation to advise for sells or buys. However, price correlations remain largely unexplored. We shed some light on the trend correlations across a large variety of cryptocoins, by investigating their coin/price correlation trends over the past two years. We study the causality between the trends, and exploit the derived correlations to understand the accuracy of state-of-the-art forecasting techniques for time series modeling (e.g., GBMs, LSTM and GRU) of correlated cryptocoins. Our evaluation shows (i) strong correlation patterns between the most traded coins (e.g., Bitcoin and Ether) and other types of cryptocurrencies, and (ii) state-of-the-art time series forecasting algorithms can be used to forecast cryptocoins price trends. We released datasets and code to reproduce our analysis to the research community.
{"title":"Practical Forecasting of Cryptocoins Timeseries using Correlation Patterns","authors":"Pasquale De Rosa, Pascal Felber, Valerio Schiavoni","doi":"arxiv-2409.03674","DOIUrl":"https://doi.org/arxiv-2409.03674","url":null,"abstract":"Cryptocoins (i.e., Bitcoin, Ether, Litecoin) are tradable digital assets.\u0000Ownerships of cryptocoins are registered on distributed ledgers (i.e.,\u0000blockchains). Secure encryption techniques guarantee the security of the\u0000transactions (transfers of coins among owners), registered into the ledger.\u0000Cryptocoins are exchanged for specific trading prices. The extreme volatility\u0000of such trading prices across all different sets of crypto-assets remains\u0000undisputed. However, the relations between the trading prices across different\u0000cryptocoins remains largely unexplored. Major coin exchanges indicate trend\u0000correlation to advise for sells or buys. However, price correlations remain\u0000largely unexplored. We shed some light on the trend correlations across a large\u0000variety of cryptocoins, by investigating their coin/price correlation trends\u0000over the past two years. We study the causality between the trends, and exploit\u0000the derived correlations to understand the accuracy of state-of-the-art\u0000forecasting techniques for time series modeling (e.g., GBMs, LSTM and GRU) of\u0000correlated cryptocoins. Our evaluation shows (i) strong correlation patterns\u0000between the most traded coins (e.g., Bitcoin and Ether) and other types of\u0000cryptocurrencies, and (ii) state-of-the-art time series forecasting algorithms\u0000can be used to forecast cryptocoins price trends. We released datasets and code\u0000to reproduce our analysis to the research community.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227895","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}
Rafael Pastrana, Eder Medina, Isabel M. de Oliveira, Sigrid Adriaenssens, Ryan P. Adams
Designing mechanically efficient geometry for architectural structures like shells, towers, and bridges is an expensive iterative process. Existing techniques for solving such inverse mechanical problems rely on traditional direct optimization methods, which are slow and computationally expensive, limiting iteration speed and design exploration. Neural networks would seem to offer an alternative, via data-driven amortized optimization for specific design tasks, but they often require extensive regularization and cannot ensure that important design criteria, such as mechanical integrity, are met. In this work, we combine neural networks with a differentiable mechanics simulator and develop a model that accelerates the solution of shape approximation problems for architectural structures. This approach allows a neural network to capture the physics of the task directly from the simulation during training, instead of having to discern it from input data and penalty terms in a physics-informed loss function. As a result, we can generate feasible designs on a variety of structural types that satisfy mechanical and geometric constraints a priori, with better accuracy than fully neural alternatives trained with handcrafted losses, while achieving comparable performance to direct optimization, but in real time. We validate our method in two distinct structural shape-matching tasks, the design of masonry shells and cable-net towers, and showcase its real-world potential for design exploration by deploying it as a plugin in commercial 3D modeling software. Our work opens up new opportunities for real-time design enhanced by neural networks of mechanically sound and efficient architectural structures in the built environment.
{"title":"Real-time design of architectural structures with differentiable simulators and neural networks","authors":"Rafael Pastrana, Eder Medina, Isabel M. de Oliveira, Sigrid Adriaenssens, Ryan P. Adams","doi":"arxiv-2409.02606","DOIUrl":"https://doi.org/arxiv-2409.02606","url":null,"abstract":"Designing mechanically efficient geometry for architectural structures like\u0000shells, towers, and bridges is an expensive iterative process. Existing\u0000techniques for solving such inverse mechanical problems rely on traditional\u0000direct optimization methods, which are slow and computationally expensive,\u0000limiting iteration speed and design exploration. Neural networks would seem to\u0000offer an alternative, via data-driven amortized optimization for specific\u0000design tasks, but they often require extensive regularization and cannot ensure\u0000that important design criteria, such as mechanical integrity, are met. In this\u0000work, we combine neural networks with a differentiable mechanics simulator and\u0000develop a model that accelerates the solution of shape approximation problems\u0000for architectural structures. This approach allows a neural network to capture\u0000the physics of the task directly from the simulation during training, instead\u0000of having to discern it from input data and penalty terms in a physics-informed\u0000loss function. As a result, we can generate feasible designs on a variety of\u0000structural types that satisfy mechanical and geometric constraints a priori,\u0000with better accuracy than fully neural alternatives trained with handcrafted\u0000losses, while achieving comparable performance to direct optimization, but in\u0000real time. We validate our method in two distinct structural shape-matching\u0000tasks, the design of masonry shells and cable-net towers, and showcase its\u0000real-world potential for design exploration by deploying it as a plugin in\u0000commercial 3D modeling software. Our work opens up new opportunities for\u0000real-time design enhanced by neural networks of mechanically sound and\u0000efficient architectural structures in the built environment.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227897","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}
Hongrui Chen, Aditya Joglekar, Zack Rubinstein, Bradley Schmerl, Gary Fedder, Jan de Nijs, David Garlan, Stephen Smith, Levent Burak Kara
Advances in CAD and CAM have enabled engineers and design teams to digitally design parts with unprecedented ease. Software solutions now come with a range of modules for optimizing designs for performance requirements, generating instructions for manufacturing, and digitally tracking the entire process from design to procurement in the form of product life-cycle management tools. However, existing solutions force design teams and corporations to take a primarily serial approach where manufacturing and procurement decisions are largely contingent on design, rather than being an integral part of the design process. In this work, we propose a new approach to part making where design, manufacturing, and supply chain requirements and resources can be jointly considered and optimized. We present the Generative Manufacturing compiler that accepts as input the following: 1) An engineering part requirements specification that includes quantities such as loads, domain envelope, mass, and compliance, 2) A business part requirements specification that includes production volume, cost, and lead time, 3) Contextual knowledge about the current manufacturing state such as availability of relevant manufacturing equipment, materials, and workforce, both locally and through the supply chain. Based on these factors, the compiler generates and evaluates manufacturing process alternatives and the optimal derivative designs that are implied by each process, and enables a user guided iterative exploration of the design space. As part of our initial implementation of this compiler, we demonstrate the effectiveness of our approach on examples of a cantilever beam problem and a rocket engine mount problem and showcase its utility in creating and selecting optimal solutions according to the requirements and resources.
{"title":"Generative Manufacturing: A requirements and resource-driven approach to part making","authors":"Hongrui Chen, Aditya Joglekar, Zack Rubinstein, Bradley Schmerl, Gary Fedder, Jan de Nijs, David Garlan, Stephen Smith, Levent Burak Kara","doi":"arxiv-2409.03089","DOIUrl":"https://doi.org/arxiv-2409.03089","url":null,"abstract":"Advances in CAD and CAM have enabled engineers and design teams to digitally\u0000design parts with unprecedented ease. Software solutions now come with a range\u0000of modules for optimizing designs for performance requirements, generating\u0000instructions for manufacturing, and digitally tracking the entire process from\u0000design to procurement in the form of product life-cycle management tools.\u0000However, existing solutions force design teams and corporations to take a\u0000primarily serial approach where manufacturing and procurement decisions are\u0000largely contingent on design, rather than being an integral part of the design\u0000process. In this work, we propose a new approach to part making where design,\u0000manufacturing, and supply chain requirements and resources can be jointly\u0000considered and optimized. We present the Generative Manufacturing compiler that\u0000accepts as input the following: 1) An engineering part requirements\u0000specification that includes quantities such as loads, domain envelope, mass,\u0000and compliance, 2) A business part requirements specification that includes\u0000production volume, cost, and lead time, 3) Contextual knowledge about the\u0000current manufacturing state such as availability of relevant manufacturing\u0000equipment, materials, and workforce, both locally and through the supply chain.\u0000Based on these factors, the compiler generates and evaluates manufacturing\u0000process alternatives and the optimal derivative designs that are implied by\u0000each process, and enables a user guided iterative exploration of the design\u0000space. As part of our initial implementation of this compiler, we demonstrate\u0000the effectiveness of our approach on examples of a cantilever beam problem and\u0000a rocket engine mount problem and showcase its utility in creating and\u0000selecting optimal solutions according to the requirements and resources.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227839","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}
Recurrent Neural Networks (RNNs) have emerged as an interesting alternative to conventional material modeling approaches, particularly for nonlinear path dependent materials. Remarkable computational enhancements are obtained using RNNs compared to classical approaches such as the computational homogenization method. However, RNN predictive errors accumulate, leading to issues when predicting temporal dependencies in time series data. This study aims to address and mitigate inaccuracies induced by neural networks in predicting path dependent plastic deformations of short fiber reinforced composite materials. We propose using an approach of Test Time data Augmentation (TTA), which, to the best of the authors knowledge, is previously untested in the context of RNNs. The method is based on augmenting the input test data using random rotations and subsequently rotating back the predicted output signal. By aggregating the back rotated predictions, a more accurate prediction compared to individual predictions is obtained. Our analysis also demonstrates improved shape consistency between the prediction and the target pseudo time signal. Additionally, this method provides an uncertainty estimation which correlates with the absolute prediction error. The TTA approach is reproducible with different randomly generated data augmentations, establishing a promising framework for optimizing predictions of deep learning models. We believe there are broader implications of the proposed method for various fields reliant on accurate predictive data driven modeling.
{"title":"Test-time data augmentation: improving predictions of recurrent neural network models of composites","authors":"Petter Uvdal, Mohsen Mirkhalaf","doi":"arxiv-2409.02478","DOIUrl":"https://doi.org/arxiv-2409.02478","url":null,"abstract":"Recurrent Neural Networks (RNNs) have emerged as an interesting alternative\u0000to conventional material modeling approaches, particularly for nonlinear path\u0000dependent materials. Remarkable computational enhancements are obtained using\u0000RNNs compared to classical approaches such as the computational homogenization\u0000method. However, RNN predictive errors accumulate, leading to issues when\u0000predicting temporal dependencies in time series data. This study aims to\u0000address and mitigate inaccuracies induced by neural networks in predicting path\u0000dependent plastic deformations of short fiber reinforced composite materials.\u0000We propose using an approach of Test Time data Augmentation (TTA), which, to\u0000the best of the authors knowledge, is previously untested in the context of\u0000RNNs. The method is based on augmenting the input test data using random\u0000rotations and subsequently rotating back the predicted output signal. By\u0000aggregating the back rotated predictions, a more accurate prediction compared\u0000to individual predictions is obtained. Our analysis also demonstrates improved\u0000shape consistency between the prediction and the target pseudo time signal.\u0000Additionally, this method provides an uncertainty estimation which correlates\u0000with the absolute prediction error. The TTA approach is reproducible with\u0000different randomly generated data augmentations, establishing a promising\u0000framework for optimizing predictions of deep learning models. We believe there\u0000are broader implications of the proposed method for various fields reliant on\u0000accurate predictive data driven modeling.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227899","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}
Moein Shahiki Tash, Zahra Ahani, Mohim Tash, Olga Kolesnikova, Grigori Sidorov
This study performs analysis of Predictive statements, Hope speech, and Regret Detection behaviors within cryptocurrency-related discussions, leveraging advanced natural language processing techniques. We introduce a novel classification scheme named "Prediction statements," categorizing comments into Predictive Incremental, Predictive Decremental, Predictive Neutral, or Non-Predictive categories. Employing GPT-4o, a cutting-edge large language model, we explore sentiment dynamics across five prominent cryptocurrencies: Cardano, Binance, Matic, Fantom, and Ripple. Our analysis reveals distinct patterns in predictive sentiments, with Matic demonstrating a notably higher propensity for optimistic predictions. Additionally, we investigate hope and regret sentiments, uncovering nuanced interplay between these emotions and predictive behaviors. Despite encountering limitations related to data volume and resource availability, our study reports valuable discoveries concerning investor behavior and sentiment trends within the cryptocurrency market, informing strategic decision-making and future research endeavors.
{"title":"Exploring Sentiment Dynamics and Predictive Behaviors in Cryptocurrency Discussions by Few-Shot Learning with Large Language Models","authors":"Moein Shahiki Tash, Zahra Ahani, Mohim Tash, Olga Kolesnikova, Grigori Sidorov","doi":"arxiv-2409.02836","DOIUrl":"https://doi.org/arxiv-2409.02836","url":null,"abstract":"This study performs analysis of Predictive statements, Hope speech, and\u0000Regret Detection behaviors within cryptocurrency-related discussions,\u0000leveraging advanced natural language processing techniques. We introduce a\u0000novel classification scheme named \"Prediction statements,\" categorizing\u0000comments into Predictive Incremental, Predictive Decremental, Predictive\u0000Neutral, or Non-Predictive categories. Employing GPT-4o, a cutting-edge large\u0000language model, we explore sentiment dynamics across five prominent\u0000cryptocurrencies: Cardano, Binance, Matic, Fantom, and Ripple. Our analysis\u0000reveals distinct patterns in predictive sentiments, with Matic demonstrating a\u0000notably higher propensity for optimistic predictions. Additionally, we\u0000investigate hope and regret sentiments, uncovering nuanced interplay between\u0000these emotions and predictive behaviors. Despite encountering limitations\u0000related to data volume and resource availability, our study reports valuable\u0000discoveries concerning investor behavior and sentiment trends within the\u0000cryptocurrency market, informing strategic decision-making and future research\u0000endeavors.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211229","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}
Generation of VLSI layout patterns is essential for a wide range of Design For Manufacturability (DFM) studies. In this study, we investigate the potential of generative machine learning models for creating design rule legal metal layout patterns. Our results demonstrate that the proposed model can generate legal patterns in complex design rule settings and achieves a high diversity score. The designed system, with its flexible settings, supports both pattern generation with localized changes, and design rule violation correction. Our methodology is validated on Intel 18A Process Design Kit (PDK) and can produce a wide range of DRC-compliant pattern libraries with only 20 starter patterns.
{"title":"PatternPaint: Generating Layout Patterns Using Generative AI and Inpainting Techniques","authors":"Guanglei Zhou, Bhargav Korrapati, Gaurav Rajavendra Reddy, Jiang Hu, Yiran Chen, Dipto G. Thakurta","doi":"arxiv-2409.01348","DOIUrl":"https://doi.org/arxiv-2409.01348","url":null,"abstract":"Generation of VLSI layout patterns is essential for a wide range of Design\u0000For Manufacturability (DFM) studies. In this study, we investigate the\u0000potential of generative machine learning models for creating design rule legal\u0000metal layout patterns. Our results demonstrate that the proposed model can\u0000generate legal patterns in complex design rule settings and achieves a high\u0000diversity score. The designed system, with its flexible settings, supports both\u0000pattern generation with localized changes, and design rule violation\u0000correction. Our methodology is validated on Intel 18A Process Design Kit (PDK)\u0000and can produce a wide range of DRC-compliant pattern libraries with only 20\u0000starter patterns.","PeriodicalId":501309,"journal":{"name":"arXiv - CS - Computational Engineering, Finance, and Science","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142211234","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：