Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face r...Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face recognition models based on convolutional neural networks.The proposed method identifies and removes inefficient filters based on the information volume in feature maps.In each layer,some feature maps lack useful information,and there exists a correlation between certain feature maps.Filters associated with these two types of feature maps impose additional computational costs on the model.By eliminating filters related to these categories of feature maps,the reduction of both computational cost and model size can be achieved.The approach employs a combination of correlation analysis and the summation of matrix elements within each feature map to detect and eliminate inefficient filters.The method was applied to two face recognition models utilizing the VGG16 and ResNet50V2 backbone architectures.In the proposed approach,the number of filters removed in each layer varies,and the removal process is independent of the adjacent layers.The convolutional layers of both backbone models were initialized with pre-trained weights from ImageNet.For training,the CASIA-WebFace dataset was utilized,and the Labeled Faces in the Wild(LFW)dataset was employed for benchmarking purposes.In the VGG16-based face recognition model,a 0.74%accuracy improvement was achieved while reducing the number of convolution parameters by 26.85%and decreasing Floating-point operations per second(FLOPs)by 47.96%.For the face recognition model based on the ResNet50V2 architecture,the ArcFace method was implemented.The removal of inactive filters in this model led to a slight decrease in accuracy by 0.11%.However,it resulted in enhanced training speed,a reduction of 59.38%in convolution parameters,and a 57.29%decrease in FLOPs.展开更多
We have investigated the resonant propagation of femtosecond laser pulse in 4-trans-[p-（N, N-Di-n-butylamino）- p-stilbenyl vinyl] pyridine medium with permanent dipole moments. The electronic structures and paramete...We have investigated the resonant propagation of femtosecond laser pulse in 4-trans-[p-（N, N-Di-n-butylamino）- p-stilbenyl vinyl] pyridine medium with permanent dipole moments. The electronic structures and parameters for the compound have been calculated by using density functional theory. In the optical regime, there is one charge-transfer state, and the molecule can thus be simplified as a two-level system. Both the one- and two-photon transitions occur between the ground and charge-transfer states. The numerical results show that the permanent dipole moments have an obvious effect on the propagation of the ultrashort pulse laser. The ideal self-induced transparency disappears for 2π pulse, and second harmonic spectral components occur significantly due to the two-photon absorption process. For the 6π pulse, continuum frequency generation is produced and a shorter duration pulse in time domain with 465 as is obtained.展开更多
Digital design of a digital signal processor involves accurate and high-speed mathematical computation units.DSP units are one of the most power consuming and memory occupying devices.Multipliers are the common buildi...Digital design of a digital signal processor involves accurate and high-speed mathematical computation units.DSP units are one of the most power consuming and memory occupying devices.Multipliers are the common building blocks in most of the DSP units which demands low power and area constraints in the field of portable biomedical devices.This research works attempts multiple power reduction technique to limit the power dissipation of the proposed LUT multiplier unit.A lookup table-based multiplier has the advantage of almost constant area requirement’s irrespective to the increase in bit size of multiplier.Clock gating is usually used to reduce the unnecessary switching activities in idle circlet components.A clock tree structure is employed to enhance the SRAM based lookup table memory architecture.The LUT memory access operation is sequential in nature and instead of address decoder a ring counter is used to scan the memory contents and gated driver tree structure is implemented to control the clock and data switching activities.The proposed algorithm yields 20%of power reduction than existing.展开更多
The correlation of surface impurity states with the antiferromagnetic ground states is crucial for understanding the formation of the topological surface state in the antiferromagnetic topological insulators MnBi_(2)T...The correlation of surface impurity states with the antiferromagnetic ground states is crucial for understanding the formation of the topological surface state in the antiferromagnetic topological insulators MnBi_(2)Te_(4).By using low-temperature scanning tunneling microscopy and spectroscopy,we observed a localized bound state around the Mn-Bi antisite defect at the Teterminated surface of the antiferromagnetic topological insulator MnBi_(2)Te_(4).When applying a magnetic field perpendicular to the surface(Bz)from–1.5 to 3.0 T,the bound state shifts linearly to a lower energy with increasing Bz,which is attributed to the Zeeman effect.Remarkably,when applying a large range of Bz from–8.0 to 8.0 T,the magnetic field induced reorientation of surface magnetic moments results in an abrupt jump in the local density of states(LDOS),which is characterized by LDOSchange-ratio■quantitatively.Interestingly,two asymmetric critical field,–2.0 and 4.0 T determined by the two peaks in■are observed,which is consistent with simulated results according to a Mills-model,describing a surface spin flop transition(SSF).Our results provide a new flatform for studying the interplay between magnetic order and topological phases in magnetic topological materials.展开更多
This is a survey paper about a selection of results in complex algebraic geometry that appeared in the recent and less recent litterature,and in which rational homogeneous spaces play a prominent role.This selection i...This is a survey paper about a selection of results in complex algebraic geometry that appeared in the recent and less recent litterature,and in which rational homogeneous spaces play a prominent role.This selection is largely arbitrary and mainly refiects the interests of the author.展开更多
文摘Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face recognition models based on convolutional neural networks.The proposed method identifies and removes inefficient filters based on the information volume in feature maps.In each layer,some feature maps lack useful information,and there exists a correlation between certain feature maps.Filters associated with these two types of feature maps impose additional computational costs on the model.By eliminating filters related to these categories of feature maps,the reduction of both computational cost and model size can be achieved.The approach employs a combination of correlation analysis and the summation of matrix elements within each feature map to detect and eliminate inefficient filters.The method was applied to two face recognition models utilizing the VGG16 and ResNet50V2 backbone architectures.In the proposed approach,the number of filters removed in each layer varies,and the removal process is independent of the adjacent layers.The convolutional layers of both backbone models were initialized with pre-trained weights from ImageNet.For training,the CASIA-WebFace dataset was utilized,and the Labeled Faces in the Wild(LFW)dataset was employed for benchmarking purposes.In the VGG16-based face recognition model,a 0.74%accuracy improvement was achieved while reducing the number of convolution parameters by 26.85%and decreasing Floating-point operations per second(FLOPs)by 47.96%.For the face recognition model based on the ResNet50V2 architecture,the ArcFace method was implemented.The removal of inactive filters in this model led to a slight decrease in accuracy by 0.11%.However,it resulted in enhanced training speed,a reduction of 59.38%in convolution parameters,and a 57.29%decrease in FLOPs.
基金Project supported by the Shandong Natural Science Foundation （Grant No Y2004A08）, the University Doctoral Subject Special Science and Technology Foundation （Grant No 20040445001） and The Key Laboratory for High Intensity 0ptics of Shanghai Institute of 0ptics and Fine Mechanics, Chinese Academy of Sciences.
文摘We have investigated the resonant propagation of femtosecond laser pulse in 4-trans-[p-（N, N-Di-n-butylamino）- p-stilbenyl vinyl] pyridine medium with permanent dipole moments. The electronic structures and parameters for the compound have been calculated by using density functional theory. In the optical regime, there is one charge-transfer state, and the molecule can thus be simplified as a two-level system. Both the one- and two-photon transitions occur between the ground and charge-transfer states. The numerical results show that the permanent dipole moments have an obvious effect on the propagation of the ultrashort pulse laser. The ideal self-induced transparency disappears for 2π pulse, and second harmonic spectral components occur significantly due to the two-photon absorption process. For the 6π pulse, continuum frequency generation is produced and a shorter duration pulse in time domain with 465 as is obtained.
文摘Digital design of a digital signal processor involves accurate and high-speed mathematical computation units.DSP units are one of the most power consuming and memory occupying devices.Multipliers are the common building blocks in most of the DSP units which demands low power and area constraints in the field of portable biomedical devices.This research works attempts multiple power reduction technique to limit the power dissipation of the proposed LUT multiplier unit.A lookup table-based multiplier has the advantage of almost constant area requirement’s irrespective to the increase in bit size of multiplier.Clock gating is usually used to reduce the unnecessary switching activities in idle circlet components.A clock tree structure is employed to enhance the SRAM based lookup table memory architecture.The LUT memory access operation is sequential in nature and instead of address decoder a ring counter is used to scan the memory contents and gated driver tree structure is implemented to control the clock and data switching activities.The proposed algorithm yields 20%of power reduction than existing.
基金This work is supported by the National Natural Science Foundation of China(Nos.61888102 and 52022105)National Key Research and Development Projects of China(Nos.2018YFA0305800 and 2019YFA0308500)+2 种基金the Strategic Priority Research Program of Chinese Academy of Sciences(Nos.XDB30000000 and XDB28000000)CAS Project for Young Scientists in Basic Research(No.YSBR-003)the University of Chinese Academy of Sciences.
文摘The correlation of surface impurity states with the antiferromagnetic ground states is crucial for understanding the formation of the topological surface state in the antiferromagnetic topological insulators MnBi_(2)Te_(4).By using low-temperature scanning tunneling microscopy and spectroscopy,we observed a localized bound state around the Mn-Bi antisite defect at the Teterminated surface of the antiferromagnetic topological insulator MnBi_(2)Te_(4).When applying a magnetic field perpendicular to the surface(Bz)from–1.5 to 3.0 T,the bound state shifts linearly to a lower energy with increasing Bz,which is attributed to the Zeeman effect.Remarkably,when applying a large range of Bz from–8.0 to 8.0 T,the magnetic field induced reorientation of surface magnetic moments results in an abrupt jump in the local density of states(LDOS),which is characterized by LDOSchange-ratio■quantitatively.Interestingly,two asymmetric critical field,–2.0 and 4.0 T determined by the two peaks in■are observed,which is consistent with simulated results according to a Mills-model,describing a surface spin flop transition(SSF).Our results provide a new flatform for studying the interplay between magnetic order and topological phases in magnetic topological materials.
文摘This is a survey paper about a selection of results in complex algebraic geometry that appeared in the recent and less recent litterature,and in which rational homogeneous spaces play a prominent role.This selection is largely arbitrary and mainly refiects the interests of the author.
