Machine Learning(ML)-based force fields are attracting ever-increasing interest due to their capacity to span spatiotemporal scales of classical interatomic potentials at quantum-level accuracy.They can be trained bas...Machine Learning(ML)-based force fields are attracting ever-increasing interest due to their capacity to span spatiotemporal scales of classical interatomic potentials at quantum-level accuracy.They can be trained based on high-fidelity simulations or experiments,the former being the common case.However,both approaches are impaired by scarce and erroneous data resulting in models that either do not agree with well-known experimental observations or are under-constrained and only reproduce some properties.Here we leverage both Density Functional Theory(DFT)calculations and experimentally measured mechanical properties and lattice parameters to train an ML potential of titanium.We demonstrate that the fused data learning strategy can concurrently satisfy all target objectives,thus resulting in a molecularmodel of higher accuracycompared to the models trained with a single data source.The inaccuracies of DFT functionals at target experimental properties were corrected,while the investigated off-target properties were affected only mildly and mostly positively.Our approach is applicable to any material and can serve as a general strategy to obtain highly accurate ML potentials.展开更多
While the forward and backward modeling of the process-structure-property chain has received a lot of attention from the materials’ community,fewer efforts have taken into consideration uncertainties.Those arise from...While the forward and backward modeling of the process-structure-property chain has received a lot of attention from the materials’ community,fewer efforts have taken into consideration uncertainties.Those arise from a multitude of sources and their quantification and integration in the inversion process are essential in meeting the materials design objectives.The first contribution of this paper is a flexible,fully probabilistic formulation of materials’ optimization problems that accounts for the uncertainty in the process-structure and structure-property linkages and enables the identification of optimal,high-dimensional,process parameters.We employ a probabilistic,data-driven surrogate for the structure-property link which expedites computations and enables handling of non-differential objectives.We couple this with a problem-tailored active learning strategy,i.e.,a self-supervised selection of training data,which significantly improves accuracy while reducing the number of expensive model simulations.We demonstrate its efficacy in optimizing the mechanical and thermal properties of two-phase,random media but envision that its applicability encompasses a wide variety of microstructure-sensitive design problems.展开更多
Neural network models for audio tasks,such as automatic speech recognition(ASR)and acoustic scene classification(ASC),are susceptible to noise contamination for real-life applications.To improve audio quality,an enhan...Neural network models for audio tasks,such as automatic speech recognition(ASR)and acoustic scene classification(ASC),are susceptible to noise contamination for real-life applications.To improve audio quality,an enhancement module,which can be developed independently,is explicitly used at the front-end of the target audio applications.In this paper,we present an end-to-end learning solution to jointly optimise the models for audio enhancement(AE)and the subsequent applications.To guide the optimisation of the AE module towards a target application,and especially to overcome difficult samples,we make use of the sample-wise performance measure as an indication of sample importance.In experiments,we consider four representative applications to evaluate our training paradigm,i.e.,ASR,speech command recognition(SCR),speech emotion recognition(SER),and ASC.These applications are associated with speech and nonspeech tasks concerning semantic and non-semantic features,transient and global information,and the experimental results indicate that our proposed approach can considerably boost the noise robustness of the models,especially at low signal-to-noise ratios,for a wide range of computer audition tasks in everyday-life noisy environments.展开更多
Predicting and monitoring battery life early and across chemistries is a significant challenge due to the plethora of degradation paths,form factors,and electrochemical testing protocols.Existing models typically tran...Predicting and monitoring battery life early and across chemistries is a significant challenge due to the plethora of degradation paths,form factors,and electrochemical testing protocols.Existing models typically translate poorly across different electrode,electrolyte,and additive materials,mostly require a fixed number of cycles,and are limited to a single discharge protocol.Here,an attention-based recurrent algorithm for neural analysis(ARCANA)architecture is developed and trained on an ultralarge,proprietary dataset from BASF and a large Li-ion dataset gathered from literature across the globe.ARCANA generalizes well across this diverse set of chemistries,electrolyte formulations,battery designs,and cycling protocols and thus allows for an extraction of data-driven knowledge of the degradation mechanisms.The model’s adaptability is further demonstrated through fine-tuning on Na-ion batteries.ARCANA advances the frontier of large-scale time series models in analytical chemistry beyond textual data and holds the potential to significantly accelerate discovery-oriented battery research endeavors.展开更多
The use of three-dimensional self-assembled metallacages(MCgs)as multimodal drug platforms holds great promise.However,the synthesis of MCgs with increased complexity and functionality is a great challenge since under...The use of three-dimensional self-assembled metallacages(MCgs)as multimodal drug platforms holds great promise.However,the synthesis of MCgs with increased complexity and functionality is a great challenge since understanding of the interaction of MCgs with biological targets is still limited.In this context,this work reports on the integration of a gold(III)porphyrin scaffold into a prismatic MCg structure and explores its application for multimodal therapy of cancer in vitro,namely enabling both photodynamic therapy and chemotherapy.Combining experimental approaches with a state-of-the-art metadynamics theoretical study,we discovered that the gold cage shows unprecedented host-vip interaction-driven selective stabilization of guanine-quadruplex(G4)structures-validated anticancer drug targets-disclosing a new mechanism to pursue in the design of supramolecular drugs.展开更多
Metal-organic frameworks(MOF)are an attractive class of porous materials due to their immense design space,allowing for application-tailored properties.Properties of interest,such as gas sorption,can be predicted in s...Metal-organic frameworks(MOF)are an attractive class of porous materials due to their immense design space,allowing for application-tailored properties.Properties of interest,such as gas sorption,can be predicted in silico with molecular mechanics simulations.However,the accuracy is limited by the available empirical force field and partial charge estimation scheme.In this work,we train a graph neural network for partial charge prediction via active learning based on Dropout Monte Carlo.We show that active learning significantly reduces the required amount of labeled MOFs to reach a target accuracy.The obtained model generalizes well to different distributions of MOFs and Zeolites.In addition,the uncertainty predictions of Dropout Monte Carlo enable reliable estimation of the mean absolute error for unseen MOFs.This work paves the way towards accurate molecular modeling of MOFs via next-generation potentials with machine learning predicted partial charges,supporting insilico material design.展开更多
Capacity planning is a very important global challenge in the face of Covid-19 pandemic.In order to hedge against the fluctuations in the random demand and to take advantage of risk pooling effect,one needs to have a ...Capacity planning is a very important global challenge in the face of Covid-19 pandemic.In order to hedge against the fluctuations in the random demand and to take advantage of risk pooling effect,one needs to have a good understanding of the variabilities in the demand of resources.However,Covid-19 predictive models that are widely used in capacity planning typically often predict the mean values of the demands(often through the predictions of the mean values of the confirmed cases and deaths)in both the temporal and spatial dimensions.They seldom provide trustworthy prediction or estimation of demand variabilities,and therefore,are insufficient for proper capacity planning.Motivated by the literature on variability scaling in the areas of physics and biology,we discovered that in the Covid-19 pandemic,both the confirmed cases and deaths exhibit a common variability scaling law between the average of the demand μ and its standard deviationσ,that is,σ ∝ μ^(β),where the scaling parameterμis typically in the range of 0.65 to 1,and the scaling law exists in both the temporal and spatial dimensions.Based on the mechanism of contagious diseases,we further build a stylized network model to explain the variability scaling phenomena.We finally provide simple models that may be used for capacity planning in both temporal and spatial dimensions,with only the predicted mean demand values from typical Covid-19 predictive models and the standard deviations of the demands derived from the variability scaling law.展开更多
基金funded by the Deutsche Forschungsgemeinschaft(DFG,German Research Foundation)-534045056.
文摘Machine Learning(ML)-based force fields are attracting ever-increasing interest due to their capacity to span spatiotemporal scales of classical interatomic potentials at quantum-level accuracy.They can be trained based on high-fidelity simulations or experiments,the former being the common case.However,both approaches are impaired by scarce and erroneous data resulting in models that either do not agree with well-known experimental observations or are under-constrained and only reproduce some properties.Here we leverage both Density Functional Theory(DFT)calculations and experimentally measured mechanical properties and lattice parameters to train an ML potential of titanium.We demonstrate that the fused data learning strategy can concurrently satisfy all target objectives,thus resulting in a molecularmodel of higher accuracycompared to the models trained with a single data source.The inaccuracies of DFT functionals at target experimental properties were corrected,while the investigated off-target properties were affected only mildly and mostly positively.Our approach is applicable to any material and can serve as a general strategy to obtain highly accurate ML potentials.
基金Funded under the Excellence Strategy of the Federal Government and the Länder in the context of the ARTEMIS Innovation Network.
文摘While the forward and backward modeling of the process-structure-property chain has received a lot of attention from the materials’ community,fewer efforts have taken into consideration uncertainties.Those arise from a multitude of sources and their quantification and integration in the inversion process are essential in meeting the materials design objectives.The first contribution of this paper is a flexible,fully probabilistic formulation of materials’ optimization problems that accounts for the uncertainty in the process-structure and structure-property linkages and enables the identification of optimal,high-dimensional,process parameters.We employ a probabilistic,data-driven surrogate for the structure-property link which expedites computations and enables handling of non-differential objectives.We couple this with a problem-tailored active learning strategy,i.e.,a self-supervised selection of training data,which significantly improves accuracy while reducing the number of expensive model simulations.We demonstrate its efficacy in optimizing the mechanical and thermal properties of two-phase,random media but envision that its applicability encompasses a wide variety of microstructure-sensitive design problems.
基金supported by the Affective Computing&HCI Innovation Research Lab between Huawei Technologies and the University of Augsburg,and the EU H2020 Project under Grant No.101135556(INDUX-R).
文摘Neural network models for audio tasks,such as automatic speech recognition(ASR)and acoustic scene classification(ASC),are susceptible to noise contamination for real-life applications.To improve audio quality,an enhancement module,which can be developed independently,is explicitly used at the front-end of the target audio applications.In this paper,we present an end-to-end learning solution to jointly optimise the models for audio enhancement(AE)and the subsequent applications.To guide the optimisation of the AE module towards a target application,and especially to overcome difficult samples,we make use of the sample-wise performance measure as an indication of sample importance.In experiments,we consider four representative applications to evaluate our training paradigm,i.e.,ASR,speech command recognition(SCR),speech emotion recognition(SER),and ASC.These applications are associated with speech and nonspeech tasks concerning semantic and non-semantic features,transient and global information,and the experimental results indicate that our proposed approach can considerably boost the noise robustness of the models,especially at low signal-to-noise ratios,for a wide range of computer audition tasks in everyday-life noisy environments.
基金This work contributes to the research performed at CELEST(Center for Electrochemical Energy Storage Ulm-Karlsruhe)and was partly funded by the German Research Foundation(DFG)under Project ID 390874152(POLiS Cluster of Excellence)This project also received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No.957189(BIG-MAP)+1 种基金funded by the European Union’s Horizon 2020 research and innovation program under Grant Agreement No.957213HSS acknowledges funding from the German Research Foundation(DFG)under Project ID 390776260(eConversion Cluster of Excellence).
文摘Predicting and monitoring battery life early and across chemistries is a significant challenge due to the plethora of degradation paths,form factors,and electrochemical testing protocols.Existing models typically translate poorly across different electrode,electrolyte,and additive materials,mostly require a fixed number of cycles,and are limited to a single discharge protocol.Here,an attention-based recurrent algorithm for neural analysis(ARCANA)architecture is developed and trained on an ultralarge,proprietary dataset from BASF and a large Li-ion dataset gathered from literature across the globe.ARCANA generalizes well across this diverse set of chemistries,electrolyte formulations,battery designs,and cycling protocols and thus allows for an extraction of data-driven knowledge of the degradation mechanisms.The model’s adaptability is further demonstrated through fine-tuning on Na-ion batteries.ARCANA advances the frontier of large-scale time series models in analytical chemistry beyond textual data and holds the potential to significantly accelerate discovery-oriented battery research endeavors.
文摘The use of three-dimensional self-assembled metallacages(MCgs)as multimodal drug platforms holds great promise.However,the synthesis of MCgs with increased complexity and functionality is a great challenge since understanding of the interaction of MCgs with biological targets is still limited.In this context,this work reports on the integration of a gold(III)porphyrin scaffold into a prismatic MCg structure and explores its application for multimodal therapy of cancer in vitro,namely enabling both photodynamic therapy and chemotherapy.Combining experimental approaches with a state-of-the-art metadynamics theoretical study,we discovered that the gold cage shows unprecedented host-vip interaction-driven selective stabilization of guanine-quadruplex(G4)structures-validated anticancer drug targets-disclosing a new mechanism to pursue in the design of supramolecular drugs.
基金the TUM Innovation Network“Artificial Intelligence Powered Multifunctional Material Design”(ARTEMIS)funded through the German Excellence Initiative and the state of Bavariathe“EXC 2089:e-conversion”DFG-cluster of excellence[project number:390776260]。
文摘Metal-organic frameworks(MOF)are an attractive class of porous materials due to their immense design space,allowing for application-tailored properties.Properties of interest,such as gas sorption,can be predicted in silico with molecular mechanics simulations.However,the accuracy is limited by the available empirical force field and partial charge estimation scheme.In this work,we train a graph neural network for partial charge prediction via active learning based on Dropout Monte Carlo.We show that active learning significantly reduces the required amount of labeled MOFs to reach a target accuracy.The obtained model generalizes well to different distributions of MOFs and Zeolites.In addition,the uncertainty predictions of Dropout Monte Carlo enable reliable estimation of the mean absolute error for unseen MOFs.This work paves the way towards accurate molecular modeling of MOFs via next-generation potentials with machine learning predicted partial charges,supporting insilico material design.
基金This research was supported in part by the National Natural Science Foundation of China(72042015,72091211,72031006 and 71722006).
文摘Capacity planning is a very important global challenge in the face of Covid-19 pandemic.In order to hedge against the fluctuations in the random demand and to take advantage of risk pooling effect,one needs to have a good understanding of the variabilities in the demand of resources.However,Covid-19 predictive models that are widely used in capacity planning typically often predict the mean values of the demands(often through the predictions of the mean values of the confirmed cases and deaths)in both the temporal and spatial dimensions.They seldom provide trustworthy prediction or estimation of demand variabilities,and therefore,are insufficient for proper capacity planning.Motivated by the literature on variability scaling in the areas of physics and biology,we discovered that in the Covid-19 pandemic,both the confirmed cases and deaths exhibit a common variability scaling law between the average of the demand μ and its standard deviationσ,that is,σ ∝ μ^(β),where the scaling parameterμis typically in the range of 0.65 to 1,and the scaling law exists in both the temporal and spatial dimensions.Based on the mechanism of contagious diseases,we further build a stylized network model to explain the variability scaling phenomena.We finally provide simple models that may be used for capacity planning in both temporal and spatial dimensions,with only the predicted mean demand values from typical Covid-19 predictive models and the standard deviations of the demands derived from the variability scaling law.
