Neural architecture search(NAS)plays an important role in many computer vision tasks.However,the high computational cost of forward and backward propagations during the search,process restricts its practical applicati...Neural architecture search(NAS)plays an important role in many computer vision tasks.However,the high computational cost of forward and backward propagations during the search,process restricts its practical application.In this paper,we present the search process as a multi-armed bandit problem,where we take into account the uncertainty caused by the contradiction between the huge search space and limited number of trials.Bandit NAS optimizes the trade-off between exploitation and exploration for a highly efficient search.Specifically,we sampled from a set of operations in one trial,where each operation was weighted by its trial performance and a bias to allow operations with less training to be selected.We further reduced the search space by abandoning the operation with the lowest potential,significantly reducing the search cost.Experimental results on the CIFAR-10 dataset show that the resulting architecture achieves the most advanced precision with a search speed approximately two times faster than that of partially connected differentialble architecture search.On ImageNet,we attained the most advanced top-1 accuracy of 75.3%with a search time of 1.8 GPU days.展开更多
Chemical synthesis of proteins containing up to 300 amino acids may cover 30%—50%of all the proteins encountered in biomedical studies and may provide an alternate approach to the usually used recombinant expression ...Chemical synthesis of proteins containing up to 300 amino acids may cover 30%—50%of all the proteins encountered in biomedical studies and may provide an alternate approach to the usually used recombinant expression teclmology,vastly expanding the chemical space of the latter.In the present review article,we tried to survey the recent progresses made for more rapid synthesis of increasingly long peptides and more efficient ligation of multiple peptide segments.The developments of seminal methods by many research groups have greatly contributed to the recent breakthroughs in the successful total synthesis of a number of functionally important proteins,such as oligoubiquitins,bacterial GroEL/ES chaperones,and mirror-image DNA polymerases.Through these studies,a potential bottleneck has also been recognized for the chemical synthesis of large proteins,namely,how to ensure that each peptide segment from a large protein avoids unfavorable aggregation when dissolved in aqueous solution.Many new methods,such as removable backbone modification(RBM)strategy have been developed to overcome this bottleneck,while more studies need to be carried out to develop more effective and less costly methods that ultimately,may lead to fully automatable chemical synthesis of customized proteins of 300 amino acids bearing any artificial designs.展开更多
基金supported by the National Natural Science Foundation of China (Grant No.62076016)。
文摘Neural architecture search(NAS)plays an important role in many computer vision tasks.However,the high computational cost of forward and backward propagations during the search,process restricts its practical application.In this paper,we present the search process as a multi-armed bandit problem,where we take into account the uncertainty caused by the contradiction between the huge search space and limited number of trials.Bandit NAS optimizes the trade-off between exploitation and exploration for a highly efficient search.Specifically,we sampled from a set of operations in one trial,where each operation was weighted by its trial performance and a bias to allow operations with less training to be selected.We further reduced the search space by abandoning the operation with the lowest potential,significantly reducing the search cost.Experimental results on the CIFAR-10 dataset show that the resulting architecture achieves the most advanced precision with a search speed approximately two times faster than that of partially connected differentialble architecture search.On ImageNet,we attained the most advanced top-1 accuracy of 75.3%with a search time of 1.8 GPU days.
基金Supported by the National Key R&D Program of China(No.2017YFA0505200)the National Natural Science Foundation of China(Nos.21532004,91753205,81621002).
文摘Chemical synthesis of proteins containing up to 300 amino acids may cover 30%—50%of all the proteins encountered in biomedical studies and may provide an alternate approach to the usually used recombinant expression teclmology,vastly expanding the chemical space of the latter.In the present review article,we tried to survey the recent progresses made for more rapid synthesis of increasingly long peptides and more efficient ligation of multiple peptide segments.The developments of seminal methods by many research groups have greatly contributed to the recent breakthroughs in the successful total synthesis of a number of functionally important proteins,such as oligoubiquitins,bacterial GroEL/ES chaperones,and mirror-image DNA polymerases.Through these studies,a potential bottleneck has also been recognized for the chemical synthesis of large proteins,namely,how to ensure that each peptide segment from a large protein avoids unfavorable aggregation when dissolved in aqueous solution.Many new methods,such as removable backbone modification(RBM)strategy have been developed to overcome this bottleneck,while more studies need to be carried out to develop more effective and less costly methods that ultimately,may lead to fully automatable chemical synthesis of customized proteins of 300 amino acids bearing any artificial designs.
