15Cr ferrite steels are urgently required in advanced Ultra-supercritical power plants but meet design challenges in balancing excellent strength and plasticity at high temperatures.We developed a three-step learning ...15Cr ferrite steels are urgently required in advanced Ultra-supercritical power plants but meet design challenges in balancing excellent strength and plasticity at high temperatures.We developed a three-step learning strategy based on mutually driven machine learning and purposeful experiments to complete this multi-objective task.Compared with traditional adaptive learning and local-interpolation learning,this step-by-step modular manner provides good transparency and interpretability of the information flow,which is ensured by identifying essential factors from an exquisitely prepared composition-microstructure dataset,and learning valuable knowledge about the composition-property relationship.The requirement of only two groups of experiments indicates the low cost and high efficiency of the strategy.Performing the strategy,we found that Ti is another key element affecting the Laves phase besides Mo and W,and their effects on ultimate tensile strength(UTS)and elongation were also uncovered.Importantly,several low-cost steels free of Co were successfully designed,and the best steel exhibited 156%,31%,and 62%higher UTS and elongation at 650°C than the typical 9Cr,15Cr,and 20Cr steels,respectively.Based on the advantages and success of the strategy in terms of alloy improvement,we believe the strategy suits other multi-objective design tasks in more materials systems.展开更多
The purpose of adversarial deep learning is to train robust DNNs against adversarial attacks,and this is one of the major research focuses of deep learning.Game theory has been used to answer some of the basic questio...The purpose of adversarial deep learning is to train robust DNNs against adversarial attacks,and this is one of the major research focuses of deep learning.Game theory has been used to answer some of the basic questions about adversarial deep learning,such as those regarding the existence of a classifier with optimal robustness and the existence of optimal adversarial samples for a given class of classifiers.In most previous works,adversarial deep learning was formulated as a simultaneous game and the strategy spaces were assumed to be certain probability distributions in order for the Nash equilibrium to exist.However,this assumption is not applicable to practical situations.In this paper,we give answers to these basic questions for the practical case where the classifiers are DNNs with a given structure;we do that by formulating adversarial deep learning in the form of Stackelberg games.The existence of Stackelberg equilibria for these games is proven.Furthermore,it is shown that the equilibrium DNN has the largest adversarial accuracy among all DNNs with the same structure,when Carlini-Wagner s margin loss is used.The trade-off between robustness and accuracy in adversarial deep learning is also studied from a game theoretical perspective.展开更多
基金supported by the National Natural Science Foundation of China(Grant Nos.51871183 and 51874245)the Research Fund of the State Key Laboratory of Solidification Processing(NPU)China(Grant No.2020-TS-06)。
文摘15Cr ferrite steels are urgently required in advanced Ultra-supercritical power plants but meet design challenges in balancing excellent strength and plasticity at high temperatures.We developed a three-step learning strategy based on mutually driven machine learning and purposeful experiments to complete this multi-objective task.Compared with traditional adaptive learning and local-interpolation learning,this step-by-step modular manner provides good transparency and interpretability of the information flow,which is ensured by identifying essential factors from an exquisitely prepared composition-microstructure dataset,and learning valuable knowledge about the composition-property relationship.The requirement of only two groups of experiments indicates the low cost and high efficiency of the strategy.Performing the strategy,we found that Ti is another key element affecting the Laves phase besides Mo and W,and their effects on ultimate tensile strength(UTS)and elongation were also uncovered.Importantly,several low-cost steels free of Co were successfully designed,and the best steel exhibited 156%,31%,and 62%higher UTS and elongation at 650°C than the typical 9Cr,15Cr,and 20Cr steels,respectively.Based on the advantages and success of the strategy in terms of alloy improvement,we believe the strategy suits other multi-objective design tasks in more materials systems.
基金This work was partially supported by NSFC(12288201)NKRDP grant(2018YFA0704705).
文摘The purpose of adversarial deep learning is to train robust DNNs against adversarial attacks,and this is one of the major research focuses of deep learning.Game theory has been used to answer some of the basic questions about adversarial deep learning,such as those regarding the existence of a classifier with optimal robustness and the existence of optimal adversarial samples for a given class of classifiers.In most previous works,adversarial deep learning was formulated as a simultaneous game and the strategy spaces were assumed to be certain probability distributions in order for the Nash equilibrium to exist.However,this assumption is not applicable to practical situations.In this paper,we give answers to these basic questions for the practical case where the classifiers are DNNs with a given structure;we do that by formulating adversarial deep learning in the form of Stackelberg games.The existence of Stackelberg equilibria for these games is proven.Furthermore,it is shown that the equilibrium DNN has the largest adversarial accuracy among all DNNs with the same structure,when Carlini-Wagner s margin loss is used.The trade-off between robustness and accuracy in adversarial deep learning is also studied from a game theoretical perspective.
