Deep Convolution Neural Networks(DCNNs)can capture discriminative features from large datasets.However,how to incrementally learn new samples without forgetting old ones and recognize novel classes that arise in the d...Deep Convolution Neural Networks(DCNNs)can capture discriminative features from large datasets.However,how to incrementally learn new samples without forgetting old ones and recognize novel classes that arise in the dynamically changing world,e.g.,classifying newly discovered fish species,remains an open problem.We address an even more challenging and realistic setting of this problem where new class samples are insufficient,i.e.,Few-Shot Class-Incremental Learning(FSCIL).Current FSCIL methods augment the training data to alleviate the overfitting of novel classes.By contrast,we propose Filter Bank Networks(FBNs)that augment the learnable filters to capture fine-detailed features for adapting to future new classes.In the forward pass,FBNs augment each convolutional filter to a virtual filter bank containing the canonical one,i.e.,itself,and multiple transformed versions.During back-propagation,FBNs explicitly stimulate fine-detailed features to emerge and collectively align all gradients of each filter bank to learn the canonical one.FBNs capture pattern variants that do not yet exist in the pretraining session,thus making it easy to incorporate new classes in the incremental learning phase.Moreover,FBNs introduce model-level prior knowledge to efficiently utilize the limited few-shot data.Extensive experiments on MNIST,CIFAR100,CUB200,andMini-ImageNet datasets show that FBNs consistently outperformthe baseline by a significantmargin,reporting new state-of-the-art FSCIL results.In addition,we contribute a challenging FSCIL benchmark,Fishshot1K,which contains 8261 underwater images covering 1000 ocean fish species.The code is included in the supplementary materials.展开更多
Self-incompatibility(SI)substantially restricts the yield and quality of citrus.Therefore,breeding and analyzing selfcompatible germplasm is of great theoretical and practical signi ficance for citrus.Here,we focus on...Self-incompatibility(SI)substantially restricts the yield and quality of citrus.Therefore,breeding and analyzing selfcompatible germplasm is of great theoretical and practical signi ficance for citrus.Here,we focus on the mechanism of a self-compatibility mutation in‘Guiyou No.1'pummelo(Citrus maxima),which is a spontaneous mutant of‘Shatian’pummelo(Citrus maxima,self-incompatibility).The rate of fruit set and the growth of pollen tubes in the pistil con firmed that a spontaneous mutation in the pistil is responsible for the self-compatibility of‘Guiyou No.1'.Segregation ratios of the S genotype in progeny,expression analysis,and western blotting validated that the reduced levels of S_(2)-RNase mRNA contribute to the loss of SI in‘Guiyou No.1'.Furthermore,we report a phased assembly of the‘Guiyou No.1'pummelo genome and obtained two complete and well-annotated S haplotypes.Coupled with an analysis of SV variations,methylation levels,and gene expression,we identi fied a candidate gene(CgHB40),that may in fluence the regulation of the S/^RNase promoter.Our data provide evidence that a mutation that affects the pistilled to the loss of SI in‘Guiyou No.1'by in fluencing a poorly understood mechanism that affects transcriptional regulation.This work signi ficantly advances our understanding of the genetic basis of the SI system in citrus and provides information on the regulation of S-RNase genes.展开更多
基金support from the Strategic Priority Research Program of the Chinese Academy of Sciences under Grant No.XDA27000000.
文摘Deep Convolution Neural Networks(DCNNs)can capture discriminative features from large datasets.However,how to incrementally learn new samples without forgetting old ones and recognize novel classes that arise in the dynamically changing world,e.g.,classifying newly discovered fish species,remains an open problem.We address an even more challenging and realistic setting of this problem where new class samples are insufficient,i.e.,Few-Shot Class-Incremental Learning(FSCIL).Current FSCIL methods augment the training data to alleviate the overfitting of novel classes.By contrast,we propose Filter Bank Networks(FBNs)that augment the learnable filters to capture fine-detailed features for adapting to future new classes.In the forward pass,FBNs augment each convolutional filter to a virtual filter bank containing the canonical one,i.e.,itself,and multiple transformed versions.During back-propagation,FBNs explicitly stimulate fine-detailed features to emerge and collectively align all gradients of each filter bank to learn the canonical one.FBNs capture pattern variants that do not yet exist in the pretraining session,thus making it easy to incorporate new classes in the incremental learning phase.Moreover,FBNs introduce model-level prior knowledge to efficiently utilize the limited few-shot data.Extensive experiments on MNIST,CIFAR100,CUB200,andMini-ImageNet datasets show that FBNs consistently outperformthe baseline by a significantmargin,reporting new state-of-the-art FSCIL results.In addition,we contribute a challenging FSCIL benchmark,Fishshot1K,which contains 8261 underwater images covering 1000 ocean fish species.The code is included in the supplementary materials.
基金This research was financially supported by the National Key Research and Development Program of China(grant no.2018YFD1000107)the National Natural Science Foundation of China(grant nos.31772259,31630065,and 31521092)+1 种基金the Fundamental Research Funds forthe Central Univer sities(grant no.2662019PY044)the China Agriculture Research System of MOF and MARA and the Hubei Provincial Natural Science Foundation of China(2020CFB532).
文摘Self-incompatibility(SI)substantially restricts the yield and quality of citrus.Therefore,breeding and analyzing selfcompatible germplasm is of great theoretical and practical signi ficance for citrus.Here,we focus on the mechanism of a self-compatibility mutation in‘Guiyou No.1'pummelo(Citrus maxima),which is a spontaneous mutant of‘Shatian’pummelo(Citrus maxima,self-incompatibility).The rate of fruit set and the growth of pollen tubes in the pistil con firmed that a spontaneous mutation in the pistil is responsible for the self-compatibility of‘Guiyou No.1'.Segregation ratios of the S genotype in progeny,expression analysis,and western blotting validated that the reduced levels of S_(2)-RNase mRNA contribute to the loss of SI in‘Guiyou No.1'.Furthermore,we report a phased assembly of the‘Guiyou No.1'pummelo genome and obtained two complete and well-annotated S haplotypes.Coupled with an analysis of SV variations,methylation levels,and gene expression,we identi fied a candidate gene(CgHB40),that may in fluence the regulation of the S/^RNase promoter.Our data provide evidence that a mutation that affects the pistilled to the loss of SI in‘Guiyou No.1'by in fluencing a poorly understood mechanism that affects transcriptional regulation.This work signi ficantly advances our understanding of the genetic basis of the SI system in citrus and provides information on the regulation of S-RNase genes.