Optimization is a concept, a process, and a method that all people use on a daily basis to solve their problems. The source of many optimization methods for many scientists has been the nature itself and the mechanism...Optimization is a concept, a process, and a method that all people use on a daily basis to solve their problems. The source of many optimization methods for many scientists has been the nature itself and the mechanisms that exist in it. Neural networks, inspired by the neurons of the human brain, have gained a great deal of recognition in recent years and provide solutions to everyday problems. Evolutionary algorithms are known for their efficiency and speed, in problems where the optimal solution is found in a huge number of possible solutions and they are also known for their simplicity, because their implementation does not require the use of complex mathematics. The combination of these two techniques is called neuroevolution. The purpose of the research is to combine and improve existing neuroevolution architectures, to solve time series problems. In this research, we propose a new improved strategy for such a system. As well as comparing the performance of our system with an already existing system, competing with it on five different datasets. Based on the final results and a combination of statistical results, we conclude that our system manages to perform much better than the existing system in all five datasets.展开更多
By combining neuroevolution potential(NEP)with phonon Boltzmann transport theory,we systematically investigate the thermal transport properties of three two-dimensional(2D)narrow bandgap semiconductors:Ca_(3)N_(2),Ba_...By combining neuroevolution potential(NEP)with phonon Boltzmann transport theory,we systematically investigate the thermal transport properties of three two-dimensional(2D)narrow bandgap semiconductors:Ca_(3)N_(2),Ba_(3)P_(2),and Ba_(3)As_(2).The room-temperature lattice thermal conductivities(κ_(L))of Ca_(3)N_(2),Ba_(3)P_(2),and Ba_(3)As_(2)considering only three-phonon scattering are 6.60 W/m K,11.90 W/m K,and 8.88 W/m K,respectively.When taking into account the higherorder phonon(four-phonon)scattering processes,theκL of these three materials decrease to 6.12 W/m K,9.73 W/m K and6.77 W/m K,respectively.Among these systems,Ba_(3)As_(2)undergoes the most pronounced suppression with a reduction of 23.8%.This is mainly due to the greater scattering phase space which enhances the four-phonon scattering.Meanwhile,it is revealed that unlike the traditional evaluation using the P_(4)/P_(3)ratio as an indicator of the strength of four-phonon interactions,the thermal conductivity of Ba_(3)P_(2)exhibits weaker four-phonon suppression behavior compared to Ba_(3)As_(2),despite hosting a higher P_(4)/P_(3)ratio.That is to say,the strength of four-phonon scattering cannot be evaluated solely by the ratio of P_(4)/P_(3).These results presented in this work shed light on the thermal transport properties of such new 2D semiconductors with narrow bandgaps.展开更多
Traditionally, heuristic re-planning algorithms are used to tackle the problem of dynamic task planning for multiple satellites. However, the traditional heuristic strategies depend on the concrete tasks, which often ...Traditionally, heuristic re-planning algorithms are used to tackle the problem of dynamic task planning for multiple satellites. However, the traditional heuristic strategies depend on the concrete tasks, which often affect the result’s optimality. Noticing that the historical information of cooperative task planning will impact the latter planning results, we propose a hybrid learning algorithm for dynamic multi-satellite task planning, which is based on the multi-agent reinforcement learning of policy iteration and the transfer learning. The reinforcement learning strategy of each satellite is described with neural networks. The policy neural network individuals with the best topological structure and weights are found by applying co-evolutionary search iteratively. To avoid the failure of the historical learning caused by the randomly occurring observation requests, a novel approach is proposed to balance the quality and efficiency of the task planning, which converts the historical learning strategy to the current initial learning strategy by applying the transfer learning algorithm. The simulations and analysis show the feasibility and adaptability of the proposed approach especially for the situation with randomly occurring observation requests.展开更多
Diamond is a wide-bandgap semiconductor with a variety of crystal configurations,and has the potential applications in the field of high-frequency,radiation-hardened,and high-power devices.There are several important ...Diamond is a wide-bandgap semiconductor with a variety of crystal configurations,and has the potential applications in the field of high-frequency,radiation-hardened,and high-power devices.There are several important polytypes of diamonds,such as cubic diamond,lonsdaleite,and nanotwinned diamond(NTD).The thermal conductivities of semiconductors in high-power devices at different temperatures should be calculated.However,there has been no reports about thermal conductivities of cubic diamond and its polytypes both efficiently and accurately based on molecular dynamics(MD).Here,using interatomic potential of neural networks can provide obvious advantages.For example,comparing with the use of density functional theory(DFT),the calculation time is reduced,while maintaining high accuracy in predicting the thermal conductivities of the above-mentioned three diamond polytypes.Based on the neuroevolution potential(NEP),the thermal conductivities of cubic diamond,lonsdaleite,and NTD at 300 K are respectively 2507.3 W·m^(-1)·K^(-1),1557.2 W·m^(-1)·K^(-1),and 985.6 W·m^(-1)·K^(-1),which are higher than the calculation results based on Tersoff-1989 potential(1508 W·m^(-1)·K^(-1),1178 W·m^(-1)·K^(-1),and 794 W·m^(-1)·K^(-1),respectively).The thermal conductivities of cubic diamond and lonsdaleite,obtained by using the NEP,are closer to the experimental data or DFT data than those from Tersoff-potential.The molecular dynamics simulations are performed by using NEP to calculate the phonon dispersions,in order to explain the possible reasons for discrepancies among the cubic diamond,lonsdaleite,and NTD.In this work,we propose a scheme to predict the thermal conductivity of cubic diamond,lonsdaleite,and NTD precisely and efficiently,and explain the differences in thermal conductivity among cubic diamond,lonsdaleite,and NTD.展开更多
In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing convention...In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing conventions problem is solved absolutely. Heuristic methods are used to constrain the topology mutation probability and the trend of mutation kind choice. Also, the niching method is used to protect the network topologies evolution. The experiment results show the efficiency and rapidity of NEGA forcefully.展开更多
Reinforcement learning is about learning agent models that make the best sequential decisions in unknown environments.In an unknown environment,the agent needs to explore the environment while exploiting the collected...Reinforcement learning is about learning agent models that make the best sequential decisions in unknown environments.In an unknown environment,the agent needs to explore the environment while exploiting the collected information,which usually forms a sophisticated problem to solve.Derivative-free optimization,meanwhile,is capable of solving sophisticated problems.It commonly uses a sampling-andupdating framework to iteratively improve the solution,where exploration and exploitation are also needed to be well balanced.Therefore,derivative-free optimization deals with a similar core issue as reinforcement learning,and has been introduced in reinforcement learning approaches,under the names of learning classifier systems and neuroevolution/evolutionary reinforcement learning.Although such methods have been developed for decades,recently,derivative-free reinforcement learning exhibits attracting increasing attention.However,recent survey on this topic is still lacking.In this article,we summarize methods of derivative-free reinforcement learning to date,and organize the methods in aspects including parameter updating,model selection,exploration,and parallel/distributed methods.Moreover,we discuss some current limitations and possible future directions,hoping that this article could bring more attentions to this topic and serve as a catalyst for developing novel and efficient approaches.展开更多
文摘Optimization is a concept, a process, and a method that all people use on a daily basis to solve their problems. The source of many optimization methods for many scientists has been the nature itself and the mechanisms that exist in it. Neural networks, inspired by the neurons of the human brain, have gained a great deal of recognition in recent years and provide solutions to everyday problems. Evolutionary algorithms are known for their efficiency and speed, in problems where the optimal solution is found in a huge number of possible solutions and they are also known for their simplicity, because their implementation does not require the use of complex mathematics. The combination of these two techniques is called neuroevolution. The purpose of the research is to combine and improve existing neuroevolution architectures, to solve time series problems. In this research, we propose a new improved strategy for such a system. As well as comparing the performance of our system with an already existing system, competing with it on five different datasets. Based on the final results and a combination of statistical results, we conclude that our system manages to perform much better than the existing system in all five datasets.
基金supported by the National Natural Science Foundation of China(Grant No.52372260)the Science Fund for Distinguished Young Scholars of Hunan Province(Grant Nos.2024JJ2048 and 2021JJ10036)+1 种基金the Science and Technology Innovation Program of Hunan Province(Grant No.2022RC1197)the Scientific Research Fund of Hunan Provincial Education Department(Grant No.22B0512)。
文摘By combining neuroevolution potential(NEP)with phonon Boltzmann transport theory,we systematically investigate the thermal transport properties of three two-dimensional(2D)narrow bandgap semiconductors:Ca_(3)N_(2),Ba_(3)P_(2),and Ba_(3)As_(2).The room-temperature lattice thermal conductivities(κ_(L))of Ca_(3)N_(2),Ba_(3)P_(2),and Ba_(3)As_(2)considering only three-phonon scattering are 6.60 W/m K,11.90 W/m K,and 8.88 W/m K,respectively.When taking into account the higherorder phonon(four-phonon)scattering processes,theκL of these three materials decrease to 6.12 W/m K,9.73 W/m K and6.77 W/m K,respectively.Among these systems,Ba_(3)As_(2)undergoes the most pronounced suppression with a reduction of 23.8%.This is mainly due to the greater scattering phase space which enhances the four-phonon scattering.Meanwhile,it is revealed that unlike the traditional evaluation using the P_(4)/P_(3)ratio as an indicator of the strength of four-phonon interactions,the thermal conductivity of Ba_(3)P_(2)exhibits weaker four-phonon suppression behavior compared to Ba_(3)As_(2),despite hosting a higher P_(4)/P_(3)ratio.That is to say,the strength of four-phonon scattering cannot be evaluated solely by the ratio of P_(4)/P_(3).These results presented in this work shed light on the thermal transport properties of such new 2D semiconductors with narrow bandgaps.
文摘Traditionally, heuristic re-planning algorithms are used to tackle the problem of dynamic task planning for multiple satellites. However, the traditional heuristic strategies depend on the concrete tasks, which often affect the result’s optimality. Noticing that the historical information of cooperative task planning will impact the latter planning results, we propose a hybrid learning algorithm for dynamic multi-satellite task planning, which is based on the multi-agent reinforcement learning of policy iteration and the transfer learning. The reinforcement learning strategy of each satellite is described with neural networks. The policy neural network individuals with the best topological structure and weights are found by applying co-evolutionary search iteratively. To avoid the failure of the historical learning caused by the randomly occurring observation requests, a novel approach is proposed to balance the quality and efficiency of the task planning, which converts the historical learning strategy to the current initial learning strategy by applying the transfer learning algorithm. The simulations and analysis show the feasibility and adaptability of the proposed approach especially for the situation with randomly occurring observation requests.
基金Project supported by the National Natural Science Foundation of China (Grant Nos.62004141 and 52202045)the Fundamental Research Funds for the Central Universities,China (Grant Nos.2042022kf1028 and 2042023kf0112)+2 种基金the Knowledge Innovation Program of Wuhan-Shuguang,China (Grant Nos.2023010201020243 and 2023010201020255)the Natural Science Foundation of Hubei Province,China (Grant No.2022CFB606)the Guangdong Basic and Applied Basic Research Fund:Guangdong–Shenzhen Joint Fund,China (Grant No.2020B1515120005)。
文摘Diamond is a wide-bandgap semiconductor with a variety of crystal configurations,and has the potential applications in the field of high-frequency,radiation-hardened,and high-power devices.There are several important polytypes of diamonds,such as cubic diamond,lonsdaleite,and nanotwinned diamond(NTD).The thermal conductivities of semiconductors in high-power devices at different temperatures should be calculated.However,there has been no reports about thermal conductivities of cubic diamond and its polytypes both efficiently and accurately based on molecular dynamics(MD).Here,using interatomic potential of neural networks can provide obvious advantages.For example,comparing with the use of density functional theory(DFT),the calculation time is reduced,while maintaining high accuracy in predicting the thermal conductivities of the above-mentioned three diamond polytypes.Based on the neuroevolution potential(NEP),the thermal conductivities of cubic diamond,lonsdaleite,and NTD at 300 K are respectively 2507.3 W·m^(-1)·K^(-1),1557.2 W·m^(-1)·K^(-1),and 985.6 W·m^(-1)·K^(-1),which are higher than the calculation results based on Tersoff-1989 potential(1508 W·m^(-1)·K^(-1),1178 W·m^(-1)·K^(-1),and 794 W·m^(-1)·K^(-1),respectively).The thermal conductivities of cubic diamond and lonsdaleite,obtained by using the NEP,are closer to the experimental data or DFT data than those from Tersoff-potential.The molecular dynamics simulations are performed by using NEP to calculate the phonon dispersions,in order to explain the possible reasons for discrepancies among the cubic diamond,lonsdaleite,and NTD.In this work,we propose a scheme to predict the thermal conductivity of cubic diamond,lonsdaleite,and NTD precisely and efficiently,and explain the differences in thermal conductivity among cubic diamond,lonsdaleite,and NTD.
文摘In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing conventions problem is solved absolutely. Heuristic methods are used to constrain the topology mutation probability and the trend of mutation kind choice. Also, the niching method is used to protect the network topologies evolution. The experiment results show the efficiency and rapidity of NEGA forcefully.
基金This work was supported by the Program A for Outstanding PhD Candidate of Nanjing University,National Science Foundation of China(61876077)Jiangsu Science Foundation(BK20170013)Collaborative Innovation Center of Novel Software Technology and Industrialization.
文摘Reinforcement learning is about learning agent models that make the best sequential decisions in unknown environments.In an unknown environment,the agent needs to explore the environment while exploiting the collected information,which usually forms a sophisticated problem to solve.Derivative-free optimization,meanwhile,is capable of solving sophisticated problems.It commonly uses a sampling-andupdating framework to iteratively improve the solution,where exploration and exploitation are also needed to be well balanced.Therefore,derivative-free optimization deals with a similar core issue as reinforcement learning,and has been introduced in reinforcement learning approaches,under the names of learning classifier systems and neuroevolution/evolutionary reinforcement learning.Although such methods have been developed for decades,recently,derivative-free reinforcement learning exhibits attracting increasing attention.However,recent survey on this topic is still lacking.In this article,we summarize methods of derivative-free reinforcement learning to date,and organize the methods in aspects including parameter updating,model selection,exploration,and parallel/distributed methods.Moreover,we discuss some current limitations and possible future directions,hoping that this article could bring more attentions to this topic and serve as a catalyst for developing novel and efficient approaches.
