Computational study of molecules and materials from first principles is a cornerstone of physics,chemistry,and materials science,but limited by the cost of accurate and precise simulations.In settings involving many s...Computational study of molecules and materials from first principles is a cornerstone of physics,chemistry,and materials science,but limited by the cost of accurate and precise simulations.In settings involving many simulations,machine learning can reduce these costs,often by orders of magnitude,by interpolating between reference simulations.This requires representations that describe any molecule or material and support interpolation.We comprehensively review and discuss current representations and relations between them.For selected state-of-the-art representations,we compare energy predictions for organic molecules,binary alloys,and Al–Ga–In sesquioxides in numerical experiments controlled for data distribution,regression method,and hyper-parameter optimization.展开更多
Global optimization of crystal compositions is a significant yet computationally intensive method to identify stable structures within chemical space.The specific physical properties linked to a threedimensional atomi...Global optimization of crystal compositions is a significant yet computationally intensive method to identify stable structures within chemical space.The specific physical properties linked to a threedimensional atomic arrangement make this an essential task in the development of new materials.We present a method that efficiently uses active learning of neural network force fields for structure relaxation,minimizing the required number of steps in the process.This is achieved by neural network force fields equipped with uncertainty estimation,which iteratively guide a pool of randomly generated candidates toward their respective local minima.Using this approach,we are able to effectively identify themost promising candidates for further evaluation using density functional theory(DFT).Our method not only reliably reduces computational costs by up to two orders of magnitude across the benchmark systemsSi_(16),Na_(8)Cl_(8),Ga_(8)As_(8)and Al_(4)O_(6)but also excels in finding themost stable minimum for the unseen,more complex systems Si46 and Al16O24.Moreover,we demonstrate at the example of Si_(16)that our method can find multiple relevant local minima while only adding minor computational effort.展开更多
Landslide detection is a hot topic in the remote sensing community,particularly with the current rapid growth in volume(and variety)of Earth observation data and the substantial progress of computer vision.Deep learni...Landslide detection is a hot topic in the remote sensing community,particularly with the current rapid growth in volume(and variety)of Earth observation data and the substantial progress of computer vision.Deep learning algorithms,especially fully convolutional networks(FCNs),and variations like the ResU-Net have been used recently as rapid and automatic landslide detection approaches.Although FCNs have shown cutting-edge results in automatic landslide detection,accuracy can be improved by adding prior knowledge through possible frameworks.This study evaluates a rulebased object-based image analysis(OBIA)approach built on probabilities resulting from the ResU-Net model for landslide detection.We train the ResU-Net model using a landslide dataset comprising landslide inventories from various geographic regions,including our study area and test the testing area not used for training.In the OBIA stage,we frst calculate land cover and image difference indices for pre-and post-landslide multi-temporal images.Next,we use the generated indices and the resulting ResU-Net probabilities for image segmentation;the extracted landslide object candidates are then optimized using rule-based classification.In the result validation section,the landslide detection of the proposed integration of the ResU-Net with a rule-based classification of OBIA is compared with that of the ResU-Net alone.Our proposed approach improves the mean intersection-over-union of the resulting map from the ResU-Net by more than 22%.展开更多
In recent years,increasingly complex machine learning methods have become state-of-the-art in modelling wind turbine power curves based on operational data.While these methods often exhibit superior performance on tes...In recent years,increasingly complex machine learning methods have become state-of-the-art in modelling wind turbine power curves based on operational data.While these methods often exhibit superior performance on test sets,they face criticism due to a perceived lack of transparency and concerns about their robustness in dynamic,non-stationary environments encountered by wind turbines.In this work,we address these issues and present a framework that leverages explainable artificial intelligence methods to gain systematic insights into data-driven power curve models.At its core,we propose a metric to quantify how well a learned model strategy aligns with the underlying physical principles of the problem.This novel tool enables model validation beyond the conventional error metrics in an automated manner.We demonstrate,for instance,its capacity as an indicator for model generalization even when limited data is available.Moreover,it facilitates understanding how decisions made during the machine learning development process,such as data selection,pre-processing,or training parameters,affect learned strategies.As a result,we obtain physically more reasonable models,a prerequisite not only for robustness but also for meaningful insights into turbine operation by domain experts.The latter,we illustrate in the context of wind turbine performance monitoring.In summary,the framework aims to guide researchers and practitioners alike toward a more informed selection and utilization of data-driven wind turbine power curve models.展开更多
基金This work received funding from the European Union’s Horizon 2020 Research and Innovation Programme,Grant Agreements No.676580,the NOMAD Laboratory CoE,and No.740233,ERC:TEC1PIt was funded in part by the German Ministry for Education and Research as BIFOLD-Berlin Institute for the Foundations of Learning and Data(ref.01IS18025A and ref.01IS18037A)Part of the research was performed while the authors visited the Institute for Pure and Applied Mathematics(IPAM),which is supported by the National Science Foundation(Grant No.DMS-1440415).
文摘Computational study of molecules and materials from first principles is a cornerstone of physics,chemistry,and materials science,but limited by the cost of accurate and precise simulations.In settings involving many simulations,machine learning can reduce these costs,often by orders of magnitude,by interpolating between reference simulations.This requires representations that describe any molecule or material and support interpolation.We comprehensively review and discuss current representations and relations between them.For selected state-of-the-art representations,we compare energy predictions for organic molecules,binary alloys,and Al–Ga–In sesquioxides in numerical experiments controlled for data distribution,regression method,and hyper-parameter optimization.
基金N.W.A.G.and M.G.contributed to this research while working at the BASLEARN-TU Berlin/BASF Joint Lab for Machine Learning,co-financed by TU Berlin and BASF SE.K.T.S.contributed to this research while working at TU Berlin and BIFOLD with grant number 01IS18037Asupported by JSPS KAKENHI Grant Number JP23H05457 and by JST-CREST Grant Number JPMJCR22O2.We thank Jonas Lederer and Klaus-Robert Müller for insightful discussions and feedback.
文摘Global optimization of crystal compositions is a significant yet computationally intensive method to identify stable structures within chemical space.The specific physical properties linked to a threedimensional atomic arrangement make this an essential task in the development of new materials.We present a method that efficiently uses active learning of neural network force fields for structure relaxation,minimizing the required number of steps in the process.This is achieved by neural network force fields equipped with uncertainty estimation,which iteratively guide a pool of randomly generated candidates toward their respective local minima.Using this approach,we are able to effectively identify themost promising candidates for further evaluation using density functional theory(DFT).Our method not only reliably reduces computational costs by up to two orders of magnitude across the benchmark systemsSi_(16),Na_(8)Cl_(8),Ga_(8)As_(8)and Al_(4)O_(6)but also excels in finding themost stable minimum for the unseen,more complex systems Si46 and Al16O24.Moreover,we demonstrate at the example of Si_(16)that our method can find multiple relevant local minima while only adding minor computational effort.
基金funded by the Institute of Advanced Research in Artificial Intelligence(IARAl)GmbHInstitute of Advanced Research in Artificial Intelligence(IARAl)GmbH Address:LandstraBer HauptstraBe 5,1030 Vienna,Austria[VAT number(UID):ATU74131236].
文摘Landslide detection is a hot topic in the remote sensing community,particularly with the current rapid growth in volume(and variety)of Earth observation data and the substantial progress of computer vision.Deep learning algorithms,especially fully convolutional networks(FCNs),and variations like the ResU-Net have been used recently as rapid and automatic landslide detection approaches.Although FCNs have shown cutting-edge results in automatic landslide detection,accuracy can be improved by adding prior knowledge through possible frameworks.This study evaluates a rulebased object-based image analysis(OBIA)approach built on probabilities resulting from the ResU-Net model for landslide detection.We train the ResU-Net model using a landslide dataset comprising landslide inventories from various geographic regions,including our study area and test the testing area not used for training.In the OBIA stage,we frst calculate land cover and image difference indices for pre-and post-landslide multi-temporal images.Next,we use the generated indices and the resulting ResU-Net probabilities for image segmentation;the extracted landslide object candidates are then optimized using rule-based classification.In the result validation section,the landslide detection of the proposed integration of the ResU-Net with a rule-based classification of OBIA is compared with that of the ResU-Net alone.Our proposed approach improves the mean intersection-over-union of the resulting map from the ResU-Net by more than 22%.
基金funded by the German Ministry for Education and Research[01IS14013A-E,01GQ1115,01GQ0850,01IS18056A,01IS18025A,and 01IS18037A]the German Research Foundation as Math+:Berlin Mathematics Research Center[EXC2046/1,project-ID:390685689]+3 种基金the Investitionsbank Berlin[10174498 ProFIT program]the European Union’s Horizon 2020 Research and Innovation program under grant[965221]funded by the Government of South Korea(MSIT)(No.2019-0-00079Artificial Intelligence Graduate School Program,Korea University and No.2022-0-00984,Development of Artificial Intelligence Technology for Personalized Plug-and-Play Explanation and Verification of Explanation).
文摘In recent years,increasingly complex machine learning methods have become state-of-the-art in modelling wind turbine power curves based on operational data.While these methods often exhibit superior performance on test sets,they face criticism due to a perceived lack of transparency and concerns about their robustness in dynamic,non-stationary environments encountered by wind turbines.In this work,we address these issues and present a framework that leverages explainable artificial intelligence methods to gain systematic insights into data-driven power curve models.At its core,we propose a metric to quantify how well a learned model strategy aligns with the underlying physical principles of the problem.This novel tool enables model validation beyond the conventional error metrics in an automated manner.We demonstrate,for instance,its capacity as an indicator for model generalization even when limited data is available.Moreover,it facilitates understanding how decisions made during the machine learning development process,such as data selection,pre-processing,or training parameters,affect learned strategies.As a result,we obtain physically more reasonable models,a prerequisite not only for robustness but also for meaningful insights into turbine operation by domain experts.The latter,we illustrate in the context of wind turbine performance monitoring.In summary,the framework aims to guide researchers and practitioners alike toward a more informed selection and utilization of data-driven wind turbine power curve models.
