Magnesium-ion batteries hold promise as future energy storage solutions,yet current Mg cathodes are challenged by low voltage and specific capacity.Herein,we present an AI-driven workflow for discovering high-performa...Magnesium-ion batteries hold promise as future energy storage solutions,yet current Mg cathodes are challenged by low voltage and specific capacity.Herein,we present an AI-driven workflow for discovering high-performance Mg cathode materials.Utilizing the common characteristics of various ionic intercalation-type electrodes,we design and train a Crystal Graph Convolutional Neural Network model that can accurately predict electrode voltages for various ions with mean absolute errors(MAE)between0.25 and 0.33 V.By deploying the trained model to stable Mg compounds from Materials Project and GNoME AI dataset,we identify 160 high voltage structures out of 15,308 candidates with voltages above3.0 V and volumetric capacity over 800 mA h/cm^(3).We further train a precise NequIP model to facilitate accurate and rapid simulations of Mg ionic conductivity.From the 160 high voltage structures,the machine learning molecular dynamics simulations have selected 23 cathode materials with both high energy density and high ionic conductivity.This Al-driven workflow dramatically boosts the efficiency and precision of material discovery for multivalent ion batteries,paving the way for advanced Mg battery development.展开更多
Very few studies have benefited from the synergetic implementation of visible,near-infrared,and shortwave infrared(VNIR-SWIR)spectra and terrain attributes in predicting Pb content in agricultural soils.To fill this g...Very few studies have benefited from the synergetic implementation of visible,near-infrared,and shortwave infrared(VNIR-SWIR)spectra and terrain attributes in predicting Pb content in agricultural soils.To fill this gap,this study aimed to predict lead(Pb)contents in agricultural soils by combining machine learning algorithms(MLAs)with VNIR-SWIR spectra or/and terrain attributes under three distinct approaches.Six MLAs were tested,including artificial neural network(ANN),partial least squares regression,support vector machine(SVM),Gaussian process regression(GPR),extreme gradient boosting(EGB),and Cubist.The VNIR-SWIR spectral data were preprocessed by methods of discrete wavelet transformation,logarithmic transformation-Savitzky Golay smoothing,standard normal variate(SNV),multiplicative scatter correction,first derivative(Fi D),and second derivative.In approach 1,MLAs were combined with the preprocessed VNIR-SWIR spectral data.The Cubist-Fi D combination was the most effective,achieving a coefficient of determination(R2)of 0.63,a concordance correlation coefficient(CCC)of 0.51,a mean absolute error(MAE)of 6.87 mg kg^(-1),and a root mean square error(RMSE)of8.66 mg kg^(-1).In approach 2,MLAs were combined with both preprocessed VNIR-SWIR spectral data and terrain attributes,and the EGB-SNV combination yielded superior results with R2of 0.75,CCC of 0.65,MAE of 5.48 mg kg^(-1),and RMSE of 7.34 mg kg^(-1).Approach 3 combined MLAs and terrain attributes,and Cubist demonstrated the best prediction results,with R^(2) of 0.75,CCC of 0.66,MAE of 6.18 mg kg^(-1),and RMSE of 7.71 mg kg^(-1).The cumulative assessment identified the fusion of terrain properties,SNV-preprocessed VNIR-SWIR spectra,and EGB as the optimal method for estimating Pb content in agricultural soils,yielding the highest R2value and minimal error.Comparatively,GPR,ANN,and SVM techniques achieved higher R2values in approaches 2 and 3 but also exhibited higher estimation errors.In conclusion,the study underscores the significance of using relevant auxiliary datasets and appropriate MLAs for accurate Pb content prediction with minimal error in agricultural soils.The findings contribute valuable insights for developing successful soil management strategies based on predictive modeling.展开更多
随着大语言模型(LLM)的快速发展,基于LLM的对话助手逐渐成为学生学习的新方式。通过学生的问答互动,对话助手能生成相应的解答,从而帮助学生解决问题,并提高学习效率。然而,现有的对话助手忽略了学生的个性化需求,无法为学生提供个性化...随着大语言模型(LLM)的快速发展,基于LLM的对话助手逐渐成为学生学习的新方式。通过学生的问答互动,对话助手能生成相应的解答,从而帮助学生解决问题,并提高学习效率。然而,现有的对话助手忽略了学生的个性化需求,无法为学生提供个性化的回答,实现“因材施教”。因此,提出一种基于学生能力感知的个性化对话助手框架。该框架包括2个主要模块:学生能力感知模块和个性化回答生成模块。能力感知模块通过分析学生的答题记录来挖掘学生的知识掌握程度,回答生成模块则根据学生的能力生成个性化回答。基于此框架,设计基于指令、基于小模型驱动和基于智能体Agent的3种实现范式,以深入探讨框架的实际效果。基于指令的对话助手利用LLM的推理能力,从学生的答题记录中挖掘知识掌握程度以帮助生成个性化回答;基于小模型驱动的对话助手利用深度知识追踪(DKT)模型生成学生的知识掌握程度;基于Agent的个性化对话助手采用LLM Agent的方式整合学生能力感知、个性化检测、答案修正等工具辅助答案的生成。基于ChatGLM(Chat General Language Model)、GPT4o_mini的对比实验结果表明,应用3种范式的LLM均能为学生提供个性化的回答,其中基于Agent的范式的准确度更高,表明该范式能更好地感知学生能力,并生成个性化回答。展开更多
基金supported by the National Key R&D Program of China(2022YFA1203400)the National Natural Science Foundation of China(W2441009)。
文摘Magnesium-ion batteries hold promise as future energy storage solutions,yet current Mg cathodes are challenged by low voltage and specific capacity.Herein,we present an AI-driven workflow for discovering high-performance Mg cathode materials.Utilizing the common characteristics of various ionic intercalation-type electrodes,we design and train a Crystal Graph Convolutional Neural Network model that can accurately predict electrode voltages for various ions with mean absolute errors(MAE)between0.25 and 0.33 V.By deploying the trained model to stable Mg compounds from Materials Project and GNoME AI dataset,we identify 160 high voltage structures out of 15,308 candidates with voltages above3.0 V and volumetric capacity over 800 mA h/cm^(3).We further train a precise NequIP model to facilitate accurate and rapid simulations of Mg ionic conductivity.From the 160 high voltage structures,the machine learning molecular dynamics simulations have selected 23 cathode materials with both high energy density and high ionic conductivity.This Al-driven workflow dramatically boosts the efficiency and precision of material discovery for multivalent ion batteries,paving the way for advanced Mg battery development.
基金supported by an institutional Ph.D.grant(No.21130/1312/3131)from the Faculty of Agrobiology,Food,and Natural Resources at the Czech University of Life Sciences Prague(CZU),Czech Republic。
文摘Very few studies have benefited from the synergetic implementation of visible,near-infrared,and shortwave infrared(VNIR-SWIR)spectra and terrain attributes in predicting Pb content in agricultural soils.To fill this gap,this study aimed to predict lead(Pb)contents in agricultural soils by combining machine learning algorithms(MLAs)with VNIR-SWIR spectra or/and terrain attributes under three distinct approaches.Six MLAs were tested,including artificial neural network(ANN),partial least squares regression,support vector machine(SVM),Gaussian process regression(GPR),extreme gradient boosting(EGB),and Cubist.The VNIR-SWIR spectral data were preprocessed by methods of discrete wavelet transformation,logarithmic transformation-Savitzky Golay smoothing,standard normal variate(SNV),multiplicative scatter correction,first derivative(Fi D),and second derivative.In approach 1,MLAs were combined with the preprocessed VNIR-SWIR spectral data.The Cubist-Fi D combination was the most effective,achieving a coefficient of determination(R2)of 0.63,a concordance correlation coefficient(CCC)of 0.51,a mean absolute error(MAE)of 6.87 mg kg^(-1),and a root mean square error(RMSE)of8.66 mg kg^(-1).In approach 2,MLAs were combined with both preprocessed VNIR-SWIR spectral data and terrain attributes,and the EGB-SNV combination yielded superior results with R2of 0.75,CCC of 0.65,MAE of 5.48 mg kg^(-1),and RMSE of 7.34 mg kg^(-1).Approach 3 combined MLAs and terrain attributes,and Cubist demonstrated the best prediction results,with R^(2) of 0.75,CCC of 0.66,MAE of 6.18 mg kg^(-1),and RMSE of 7.71 mg kg^(-1).The cumulative assessment identified the fusion of terrain properties,SNV-preprocessed VNIR-SWIR spectra,and EGB as the optimal method for estimating Pb content in agricultural soils,yielding the highest R2value and minimal error.Comparatively,GPR,ANN,and SVM techniques achieved higher R2values in approaches 2 and 3 but also exhibited higher estimation errors.In conclusion,the study underscores the significance of using relevant auxiliary datasets and appropriate MLAs for accurate Pb content prediction with minimal error in agricultural soils.The findings contribute valuable insights for developing successful soil management strategies based on predictive modeling.
文摘随着大语言模型(LLM)的快速发展,基于LLM的对话助手逐渐成为学生学习的新方式。通过学生的问答互动,对话助手能生成相应的解答,从而帮助学生解决问题,并提高学习效率。然而,现有的对话助手忽略了学生的个性化需求,无法为学生提供个性化的回答,实现“因材施教”。因此,提出一种基于学生能力感知的个性化对话助手框架。该框架包括2个主要模块:学生能力感知模块和个性化回答生成模块。能力感知模块通过分析学生的答题记录来挖掘学生的知识掌握程度,回答生成模块则根据学生的能力生成个性化回答。基于此框架,设计基于指令、基于小模型驱动和基于智能体Agent的3种实现范式,以深入探讨框架的实际效果。基于指令的对话助手利用LLM的推理能力,从学生的答题记录中挖掘知识掌握程度以帮助生成个性化回答;基于小模型驱动的对话助手利用深度知识追踪(DKT)模型生成学生的知识掌握程度;基于Agent的个性化对话助手采用LLM Agent的方式整合学生能力感知、个性化检测、答案修正等工具辅助答案的生成。基于ChatGLM(Chat General Language Model)、GPT4o_mini的对比实验结果表明,应用3种范式的LLM均能为学生提供个性化的回答,其中基于Agent的范式的准确度更高,表明该范式能更好地感知学生能力,并生成个性化回答。
