On-device Artificial Intelligence(AI)accelerators capable of not only inference but also training neural network models are in increasing demand in the industrial AI field,where frequent retraining is crucial due to f...On-device Artificial Intelligence(AI)accelerators capable of not only inference but also training neural network models are in increasing demand in the industrial AI field,where frequent retraining is crucial due to frequent production changes.Batch normalization(BN)is fundamental to training convolutional neural networks(CNNs),but its implementation in compact accelerator chips remains challenging due to computational complexity,particularly in calculating statistical parameters and gradients across mini-batches.Existing accelerator architectures either compromise the training accuracy of CNNs through approximations or require substantial computational resources,limiting their practical deployment.We present a hardware-optimized BN accelerator that maintains training accuracy while significantly reducing computational overhead through three novel techniques:(1)resourcesharing for efficient resource utilization across forward and backward passes,(2)interleaved buffering for reduced dynamic random-access memory(DRAM)access latencies,and(3)zero-skipping for minimal gradient computation.Implemented on a VCU118 Field Programmable Gate Array(FPGA)on 100 MHz and validated using You Only Look Once version 2-tiny(YOLOv2-tiny)on the PASCALVisualObjectClasses(VOC)dataset,our normalization accelerator achieves a 72%reduction in processing time and 83%lower power consumption compared to a 2.4 GHz Intel Central Processing Unit(CPU)software normalization implementation,while maintaining accuracy(0.51%mean Average Precision(mAP)drop at floating-point 32 bits(FP32),1.35%at brain floating-point 16 bits(bfloat16)).When integrated into a neural processing unit(NPU),the design demonstrates 63%and 97%performance improvements over AMD CPU and Reduced Instruction Set Computing-V(RISC-V)implementations,respectively.These results confirm that our proposed BN hardware design enables efficient,high-accuracy,and power-saving on-device training for modern CNNs.Our results demonstrate that efficient hardware implementation of standard batch normalization is achievable without sacrificing accuracy,enabling practical on-device CNN training with significantly reduced computational and power requirements.展开更多
The rapid rise of cyberattacks and the gradual failure of traditional defense systems and approaches led to using artificial intelligence(AI)techniques(such as machine learning(ML)and deep learning(DL))to build more e...The rapid rise of cyberattacks and the gradual failure of traditional defense systems and approaches led to using artificial intelligence(AI)techniques(such as machine learning(ML)and deep learning(DL))to build more efficient and reliable intrusion detection systems(IDSs).However,the advent of larger IDS datasets has negatively impacted the performance and computational complexity of AI-based IDSs.Many researchers used data preprocessing techniques such as feature selection and normalization to overcome such issues.While most of these researchers reported the success of these preprocessing techniques on a shallow level,very few studies have been performed on their effects on a wider scale.Furthermore,the performance of an IDS model is subject to not only the utilized preprocessing techniques but also the dataset and the ML/DL algorithm used,which most of the existing studies give little emphasis on.Thus,this study provides an in-depth analysis of feature selection and normalization effects on IDS models built using three IDS datasets:NSL-KDD,UNSW-NB15,and CSE–CIC–IDS2018,and various AI algorithms.A wrapper-based approach,which tends to give superior performance,and min-max normalization methods were used for feature selection and normalization,respectively.Numerous IDS models were implemented using the full and feature-selected copies of the datasets with and without normalization.The models were evaluated using popular evaluation metrics in IDS modeling,intra-and inter-model comparisons were performed between models and with state-of-the-art works.Random forest(RF)models performed better on NSL-KDD and UNSW-NB15 datasets with accuracies of 99.86%and 96.01%,respectively,whereas artificial neural network(ANN)achieved the best accuracy of 95.43%on the CSE–CIC–IDS2018 dataset.The RF models also achieved an excellent performance compared to recent works.The results show that normalization and feature selection positively affect IDS modeling.Furthermore,while feature selection benefits simpler algorithms(such as RF),normalization is more useful for complex algorithms like ANNs and deep neural networks(DNNs),and algorithms such as Naive Bayes are unsuitable for IDS modeling.The study also found that the UNSW-NB15 and CSE–CIC–IDS2018 datasets are more complex and more suitable for building and evaluating modern-day IDS than the NSL-KDD dataset.Our findings suggest that prioritizing robust algorithms like RF,alongside complex models such as ANN and DNN,can significantly enhance IDS performance.These insights provide valuable guidance for managers to develop more effective security measures by focusing on high detection rates and low false alert rates.展开更多
In recent decades,brain tumors have emerged as a serious neurological disorder that often leads to death.Hence,Brain Tumor Segmentation(BTS)is significant to enable the visualization,classification,and delineation of ...In recent decades,brain tumors have emerged as a serious neurological disorder that often leads to death.Hence,Brain Tumor Segmentation(BTS)is significant to enable the visualization,classification,and delineation of tumor regions in Magnetic Resonance Imaging(MRI).However,BTS remains a challenging task because of noise,non-uniform object texture,diverse image content and clustered objects.To address these challenges,a novel model is implemented in this research.The key objective of this research is to improve segmentation accuracy and generalization in BTS by incorporating Switchable Normalization into Faster R-CNN,which effectively captures the fine-grained tumor features to enhance segmentation precision.MRI images are initially acquired from three online datasets:Dataset 1—Brain Tumor Segmentation(BraTS)2018,Dataset 2—BraTS 2019,and Dataset 3—BraTS 2020.Subsequently,the Switchable Normalization-based Faster Regions with Convolutional Neural Networks(SNFRC)model is proposed for improved BTS in MRI images.In the proposed model,Switchable Normalization is integrated into the conventional architecture,enhancing generalization capability and reducing overfitting to unseen image data,which is essential due to the typically limited size of available datasets.The network depth is increased to obtain discriminative semantic features that improve segmentation performance.Specifically,Switchable Normalization captures the diverse feature representations from the brain images.The Faster R-CNN model develops end-to-end training and effective regional proposal generation,with an enhanced training stability using Switchable Normalization,to perform an effective segmentation in MRI images.From the experimental results,the proposed model attains segmentation accuracies of 99.41%,98.12%,and 96.71%on Datasets 1,2,and 3,respectively,outperforming conventional deep learning models used for BTS.展开更多
Renormalization group analysis has been proposed to eliminate secular terms in perturbation solutions of differential equations and thus expand the domain of their validity.Here we extend the method to treat periodic ...Renormalization group analysis has been proposed to eliminate secular terms in perturbation solutions of differential equations and thus expand the domain of their validity.Here we extend the method to treat periodic orbits or limit cycles.Interesting normal forms could be derived through a generalization of the concept'resonance',which offers nontrivial analytic approximations.Compared with traditional techniques such as multi-scale methods,the current scheme proceeds in a very straightforward and simple way,delivering not only the period and the amplitude but also the transient path to limit cycles.The method is demonstrated with several examples including the Duffing oscillator,van der Pol equation and Lorenz equation.The obtained solutions match well with numerical results and with those derived by traditional analytic methods.展开更多
Tumor vascular normalization has emerged as a promising strategy for synergistic therapy recently.Based on the strategy of“fluorescence turn on-controllable release”,a novel bifunctional candidate was con-structed b...Tumor vascular normalization has emerged as a promising strategy for synergistic therapy recently.Based on the strategy of“fluorescence turn on-controllable release”,a novel bifunctional candidate was con-structed based on previous developed vascular normalization inducer QDAU5,which could self-assemble to form functional enzyme infrared QDAU5 nanoparticles(FEIRQ NPs).Subsequently,biological evaluation demonstrated that the FEIRQ NPs could induce ferroptosis,endoplasmic reticulum stress,and antigen pre-conditioning and maturation of dendritic cells and CD8^(+)T cells,leading to excellent antitumor efficacy in the absence of cytotoxic drugs.Additionally,FEIRQ NPs show high fluorescence intensity upon expo-sure to theβ-galactosidase(β-Gal)enzyme expressed in ovarian cancer,enabling real-time monitoring of therapeutic effects.Overall,our findings suggest a prospering strategy to early diagnosis and efficient therapy for ovarian cancer without cytotoxicity.展开更多
Remodeling tumor microenvironment(TME)is a very promising and effective strategy to enhance the effects of chemotherapy,photodynamic therapy,and immunotherapy.Normalization of tumor vasculature as well as depletion of...Remodeling tumor microenvironment(TME)is a very promising and effective strategy to enhance the effects of chemotherapy,photodynamic therapy,and immunotherapy.Normalization of tumor vasculature as well as depletion of glutathione(GSH)can improve the TME.Here,we developed a novel therapeutic nanoparticle functional enzyme ultra QDAU5 nanoparticles(FEUQ Nps)based on a fluorescence-on and releasable strategy by combining a vascular normalization inducer,a GSH depleting agent,and an activated fluorophore.In which the cleavage of disulfide bonds releases active molecules that induce vascular normalization and improve the hypoxic microenvironment.In addition,it may deplete GSH in cancer cells,thus inducing the production of reactive oxygen species(ROS)and lipid peroxide(LPO)and promoting iron toxicity.It may also lead to endoplasmic stress and release of calmodulin,which activates the immune system.Meanwhile,quenched fluorophores are turned on in the presence of galactosidase(GLU)for tumor-specific labeling.In summary,we developed novel therapeutic agent nanoparticles with the function of vascular normalization inducers to achieve specific labeling of hepatocellular carcinoma while exerting efficient antitumor effects in vivo.展开更多
Due to differences in the distribution of scores for different trials, the performance of a speaker verification system will be seriously diminished if raw scores are directly used for detection with a unified thresho...Due to differences in the distribution of scores for different trials, the performance of a speaker verification system will be seriously diminished if raw scores are directly used for detection with a unified threshold value. As such, the scores must be normalized. To tackle the shortcomings of score normalization methods, we propose a speaker verification system based on log-likelihood normalization (LLN). Without a priori knowledge, LLN increases the separation between scores of target and non-target speaker models, so as to improve score aliasing of “same-speaker” and “different-speaker” trials corresponding to the same test speech, enabling better discrimination and decision capability. The experiment shows that LLN is an effective method of scoring normalization.展开更多
针对说话人确认中,各目标话者模型输出评分分布不一致而导致系统确认阈值设置的困难,本文采取了通过评分规整确定系统最小检测代价函数(DCF)确认阈值的方法。在分析了已有的两种评分规整方法Z norm a l-ization和T norm a lization的基...针对说话人确认中,各目标话者模型输出评分分布不一致而导致系统确认阈值设置的困难,本文采取了通过评分规整确定系统最小检测代价函数(DCF)确认阈值的方法。在分析了已有的两种评分规整方法Z norm a l-ization和T norm a lization的基础上,提出了一种结合两者优点的组合规整方法——TZ norm a lization,并据此给出了一种阈值动态修正方法,有效地提高了系统的性能和阈值选取的鲁棒性。对历年的N IST(手机电话语音)评测语料库进行了实验,表明了该方法的有效性。展开更多
Robust normalization is a prerequisite for reliable metabonomic analysis especially when intervention treatments cause drastic metabolomic changes or when spot urinary samples are employed without knowing the drinking...Robust normalization is a prerequisite for reliable metabonomic analysis especially when intervention treatments cause drastic metabolomic changes or when spot urinary samples are employed without knowing the drinking water quantity.With the simulated and real datasets,here,we report a probabilistic quotient normalization method based on the mode-of-quotients(mPQN)which is suitable for metabonomic analysis of both NMR and LC-MS data with little and/or drastic metabolite changes.When applied to metabonomic analysis of both animal plasma samples and human urinary samples,this newly proposed method has clearly shown better robustness than all classical normalization methods especially when drastic changes of some metabolites occur.展开更多
During the process of coal prospecting and exploration, different measurement time, different logging instruments and series can lead to systematic errors in well logs. Accordingly, all logging curves need to be norma...During the process of coal prospecting and exploration, different measurement time, different logging instruments and series can lead to systematic errors in well logs. Accordingly, all logging curves need to be normalized in the mining area. By studying well-logging normalization methods, and focusing on the characteristics of the coalfield, the frequency histogram method was used in accordance with the condition of the Guqiao Coal Mine. In this way, the density and sonic velocity at marker bed in the non-key well were made to close to those in the key well, and were eventually equal. Well log normalization was completed when this method was applied to the entire logging curves. The results show that the scales of logging data were unified by normalizing coal logging curves, and the logging data were consistent with wave impedance inversion data. A satisfactory inversion effect was obtained.展开更多
A novel approach is proposed to detect the normal vector to product surface in real time for the robotic precision drilling system in aircraft component assembly, and the auto-normalization algorithm is presented base...A novel approach is proposed to detect the normal vector to product surface in real time for the robotic precision drilling system in aircraft component assembly, and the auto-normalization algorithm is presented based on the detection system. Firstly, the deviation between the normal vector and the spindle axis is measured by the four laser displacement sensors installed at the head of the multi-function end effector. Then, the robot target attitude is inversely solved according to the auto-normalization algorithm. Finally, adjust the robot to the target attitude via pitch and yaw rotations about the tool center point and the spindle axis is corrected in line with the normal vector simultaneously. To test and verify the auto-normalization algorithm, an experimental platform is established in which the laser tracker is introduced for accurate measurement. The results show that the deviations between the corrected spindle axis and the normal vector are all reduced to less than 0.5°, with the mean value 0.32°. It is demonstrated the detection method and the autonormalization algorithm are feasible and reliable.展开更多
In order to evaluate radiometric normalization techniques, two image normalization algorithms for absolute radiometric correction of Landsat imagery were quantitatively compared in this paper, which are the Illuminati...In order to evaluate radiometric normalization techniques, two image normalization algorithms for absolute radiometric correction of Landsat imagery were quantitatively compared in this paper, which are the Illumination Correction Model proposed by Markham and Irish and the Illumination and Atmospheric Correction Model developed by the Remote Sensing and GIS Laboratory of the Utah State University. Relative noise, correlation coefficient and slope value were used as the criteria for the evaluation and comparison, which were derived from pseudo-invarlant features identified from multitemporal Landsat image pairs of Xiamen （厦门） and Fuzhou （福州） areas, both located in the eastern Fujian （福建） Province of China. Compared with the unnormalized image, the radiometric differences between the normalized multitemporal images were significantly reduced when the seasons of multitemporal images were different. However, there was no significant difference between the normalized and unnorrealized images with a similar seasonal condition. Furthermore, the correction results of two algorithms are similar when the images are relatively clear with a uniform atmospheric condition. Therefore, the radiometric normalization procedures should be carried out if the multitemporal images have a significant seasonal difference.展开更多
This study successfully deals with the inhomogeneous dimension problem of load separation assumption, which is the theoretical basis of the normalization method. According to the dimensionless load separation principl...This study successfully deals with the inhomogeneous dimension problem of load separation assumption, which is the theoretical basis of the normalization method. According to the dimensionless load separation principle, the normalization method has been improved by intro- ducing a forcible blunting correction. With the improved normalization method, the J-resistance curves of five different metallic materials of CT and SEB specimens are estimated. The forcible blunting correction of initial crack size plays an important role in the J-resistance curve estima- tion, which is closely related to the strain hardening level of material. The higher level of strain hardening leads to a greater difference in JQ determined by different slopes of the blunting line. If the blunting line coefficient recommended by ASTM E1820-11 is used in the improved nor- realization method, it will lead to greater fracture resistance than that processed by the blunting line coefficient recommended by ISO 12135-2002. Therefore, the influence of the blunting line on the determination of JQ must be taken into full account in the fracture toughness assessment of metallic materials.展开更多
针对因特网上数字图像的版权保护、认证和完整性等问题,基于DCT变换、image moment normalization和m序列,提出了一种二值水印嵌入算法,实现了二值图像的嵌入和提取。根据m序列的伪随机性和抗干扰性能,使水印具有良好的隐蔽性和稳健性;...针对因特网上数字图像的版权保护、认证和完整性等问题,基于DCT变换、image moment normalization和m序列,提出了一种二值水印嵌入算法,实现了二值图像的嵌入和提取。根据m序列的伪随机性和抗干扰性能,使水印具有良好的隐蔽性和稳健性;使用了moment normalization能抵制各种几何攻击。实验表明该算法具有很好的鲁棒性、实用性和可操作性。展开更多
There is a lack of systematic research on the expression of internal control genes used for gene expression normalization in real-time reverse transcription polymerase chain reaction in spinal cord injury research.In ...There is a lack of systematic research on the expression of internal control genes used for gene expression normalization in real-time reverse transcription polymerase chain reaction in spinal cord injury research.In this study,we used rat models of spinal cord hemisection to analyze the expression stability of 13 commonly applied reference genes:Actb,Ankrd27,CypA,Gapdh,Hprt1,Mrpl10,Pgk1,Rictor,Rn18s,Tbp,Ubc,Ubxn11,and Ywhaz.Our results show that the expression of Ankrd27,Ubc,and Tbp were stable after spinal cord injury,while Actb was the most unstable internal control gene.Ankrd27,Ubc,Tbp,and Actb were consequently used to investigate the effects of internal control genes with differing stabilities on the normalization of target gene expression.Target gene expression levels and changes over time were similar when Ankrd27,Ubc,and Tbp were used as internal controls but different when Actb was used as an internal control.We recommend that Ankrd27,Ubc,and Tbp are used as internal control genes for real-time reverse transcription polymerase chain reaction in spinal cord injury research.This study was approved by the Administration Committee of Experimental Animals,Jiangsu Province,China(approval No.20180304-008)on March 4,2018.展开更多
Radiometric normalization,as an essential step for multi-source and multi-temporal data processing,has received critical attention.Relative Radiometric Normalization(RRN)method has been primarily used for eliminating ...Radiometric normalization,as an essential step for multi-source and multi-temporal data processing,has received critical attention.Relative Radiometric Normalization(RRN)method has been primarily used for eliminating the radiometric inconsistency.The radiometric trans-forming relation between the subject image and the reference image is an essential aspect of RRN.Aimed at accurate radiometric transforming relation modeling,the learning-based nonlinear regression method,Support Vector machine Regression(SVR)is used for fitting the complicated radiometric transforming relation for the coarse-resolution data-referenced RRN.To evaluate the effectiveness of the proposed method,a series of experiments are performed,including two synthetic data experiments and one real data experiment.And the proposed method is compared with other methods that use linear regression,Artificial Neural Network(ANN)or Random Forest(RF)for radiometric transforming relation modeling.The results show that the proposed method performs well on fitting the radiometric transforming relation and could enhance the RRN performance.展开更多
Background： Expression levels for genes of interest must be normalized with an appropriate reference, or housekeeping gene, to make accurate comparisons of quantitative real-time PCR results. The purpose of this stud...Background： Expression levels for genes of interest must be normalized with an appropriate reference, or housekeeping gene, to make accurate comparisons of quantitative real-time PCR results. The purpose of this study was to identify the most stable housekeeping genes in porcine articular cartilage subjected to a mechanical injury from a panel of 10 candidate genes. Results： Ten candidate housekeeping genes were evaluated in three different treatment groups of mechanically impacted porcine articular cartilage. The genes evaluated were： beta actin, beta-2-microglobulin, glyceraldehyde-3-phosphate dehydrogenase, hydroxymethylbilane synthase, hypoxanthine phosphoribosyl transferase, peptidylprolyl isomerase A （cyclophilin A）, ribosomal protein L4, succinate dehydrogenase flavoprotein subunit A, TATA box binding protein, and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein--zeta polypeptide The stability of the genes was measured using geNorm, BestKeeper, and NormFinder software. The four most stable genes measured via geNorm were （most to least stable） succinate dehydrogenase flavoprotein, subunit A, peptidylprolyl isomerase A, glyceraldehyde-3-phosphate dehydrogenase, beta actin; the four most stable genes measured via BestKeeper were glyceraldehyde-3-phosphate dehydrogenase, peptidylprolyl isomerase A, beta actin, succinate dehydrogenase flavoprotein, subunit A; and the four most stable genes measured via NormFinder were peptidylprolyl isomerase A, sucdnate dehydrogenase flavoprotein, subunit A, glyceraldehyde-3-phosphate dehydrogenase, beta actin. Conclusions： BestKeeper, geNorm, and NormFinder all generated similar results for the most stable genes in porcine articular cartilage. The use of these appropriate reference genes will facilitate accurate gene expression studies of porcine articular cartilage and suggest appropriate housekeeping genes for articular cartilage studies in other species.展开更多
The hydrodynamic coefficients C-d and C-m are not only dependent on the size of slender cylinder, its location in water, KC number and Re number, but also vary with environmental conditions, i.e., in regular waves or ...The hydrodynamic coefficients C-d and C-m are not only dependent on the size of slender cylinder, its location in water, KC number and Re number, but also vary with environmental conditions, i.e., in regular waves or in irregular waves, in pure waves or in wave-current coexisting field. In this paper, the normalization of hydrodynamic coefficients for various environmental conditions is discussed. When a proper definition of KC number and proper characteristic values of irregular waves are used, a unified relationship between C-d, C-m and KC number for regular waves, irregular waves, pure waves and wave-current coexisting field can be obtained.展开更多
Quantitative real-time reverse transcription-polymerase chain reaction （qRT-PCR） is widely used in studies of gene expression. In most of these studies, housekeeping genes are used as internal references without val...Quantitative real-time reverse transcription-polymerase chain reaction （qRT-PCR） is widely used in studies of gene expression. In most of these studies, housekeeping genes are used as internal references without validation. To identify appropriate reference genes for qRT-PCR in Pacific abalone Haliotis discus hannai, we examined the transcription stability of six housekeeping genes in abalone tissues in the presence and absence of bacterial infection. For this purpose, abalone were infected with the bacterial pathogen Fibrio anguillarum for 12 h and 48 h. The mRNA levels of the housekeeping genes in five tissues （digestive glands, foot muscle, gill, hemocyte, and mantle） were determined by qRT-PCR. The PCR data was subsequently analyzed with the geNorm and NormFinder algorithms. The results show that in the absence of bacterial infection, elongation factor-l-alpha and beta-actin were the most stably expressed genes in all tissues, and thus are suitable as cross-tissue type normalization factors. However, we did not identify any universal reference genes post infection because the most stable genes varied between tissue types. Furthermore, for most tissues, the optimal reference genes identified by both algorithms at 12 h and 48 h post-infection differed. These results indicate that bacterial infection induced significant changes in the expression of abalone housekeeping genes in a manner that is dependent on tissue type and duration of infection. As a result, different normalization factors must be used for different tissues at different infection points.展开更多
The changes of blood perfusion and oxygen transport in tumors during tumor vascular normalization are studied with 3-dimensional mathematical modeling and numerical simulation. The models of tumor angiogenesis and vas...The changes of blood perfusion and oxygen transport in tumors during tumor vascular normalization are studied with 3-dimensional mathematical modeling and numerical simulation. The models of tumor angiogenesis and vascular-disrupting are used to simulate ＂un-normalized＂ and ＂normalized＂ vasculatures. A new model combining tumor hemodynamics and oxygen transport is developed. In this model, the intravasculartransvascular-interstitial flow with red blood cell（RBC） delivery is tightly coupled, and the oxygen resource is produced by heterogeneous distribution of hematocrit from the flow simulation. The results show that both tumor blood perfusion and hematocrit in the vessels increase, and the hypoxia microenvironment in the tumor center is greatly improved during vascular normalization. The total oxygen content inside the tumor tissue increases by about 67%, 51%, and 95% for the three approaches of vascular normalization,respectively. The elevation of oxygen concentration in tumors can improve its metabolic environment, and consequently reduce malignancy of tumor cells. It can also enhance radiation and chemotherapeutics to tumors.展开更多
基金supported by the National Research Foundation of Korea(NRF)grant for RLRC funded by the Korea government(MSIT)(No.2022R1A5A8026986,RLRC)supported by Institute of Information&Communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2020-0-01304,Development of Self-Learnable Mobile Recursive Neural Network Processor Technology)+3 种基金supported by the MSIT(Ministry of Science and ICT),Republic of Korea,under the Grand Information Technology Research Center support program(IITP-2024-2020-0-01462,Grand-ICT)supervised by the IITP(Institute for Information&Communications Technology Planning&Evaluation)supported by the Korea Technology and Information Promotion Agency for SMEs(TIPA)supported by the Korean government(Ministry of SMEs and Startups)’s Smart Manufacturing Innovation R&D(RS-2024-00434259).
文摘On-device Artificial Intelligence(AI)accelerators capable of not only inference but also training neural network models are in increasing demand in the industrial AI field,where frequent retraining is crucial due to frequent production changes.Batch normalization(BN)is fundamental to training convolutional neural networks(CNNs),but its implementation in compact accelerator chips remains challenging due to computational complexity,particularly in calculating statistical parameters and gradients across mini-batches.Existing accelerator architectures either compromise the training accuracy of CNNs through approximations or require substantial computational resources,limiting their practical deployment.We present a hardware-optimized BN accelerator that maintains training accuracy while significantly reducing computational overhead through three novel techniques:(1)resourcesharing for efficient resource utilization across forward and backward passes,(2)interleaved buffering for reduced dynamic random-access memory(DRAM)access latencies,and(3)zero-skipping for minimal gradient computation.Implemented on a VCU118 Field Programmable Gate Array(FPGA)on 100 MHz and validated using You Only Look Once version 2-tiny(YOLOv2-tiny)on the PASCALVisualObjectClasses(VOC)dataset,our normalization accelerator achieves a 72%reduction in processing time and 83%lower power consumption compared to a 2.4 GHz Intel Central Processing Unit(CPU)software normalization implementation,while maintaining accuracy(0.51%mean Average Precision(mAP)drop at floating-point 32 bits(FP32),1.35%at brain floating-point 16 bits(bfloat16)).When integrated into a neural processing unit(NPU),the design demonstrates 63%and 97%performance improvements over AMD CPU and Reduced Instruction Set Computing-V(RISC-V)implementations,respectively.These results confirm that our proposed BN hardware design enables efficient,high-accuracy,and power-saving on-device training for modern CNNs.Our results demonstrate that efficient hardware implementation of standard batch normalization is achievable without sacrificing accuracy,enabling practical on-device CNN training with significantly reduced computational and power requirements.
文摘The rapid rise of cyberattacks and the gradual failure of traditional defense systems and approaches led to using artificial intelligence(AI)techniques(such as machine learning(ML)and deep learning(DL))to build more efficient and reliable intrusion detection systems(IDSs).However,the advent of larger IDS datasets has negatively impacted the performance and computational complexity of AI-based IDSs.Many researchers used data preprocessing techniques such as feature selection and normalization to overcome such issues.While most of these researchers reported the success of these preprocessing techniques on a shallow level,very few studies have been performed on their effects on a wider scale.Furthermore,the performance of an IDS model is subject to not only the utilized preprocessing techniques but also the dataset and the ML/DL algorithm used,which most of the existing studies give little emphasis on.Thus,this study provides an in-depth analysis of feature selection and normalization effects on IDS models built using three IDS datasets:NSL-KDD,UNSW-NB15,and CSE–CIC–IDS2018,and various AI algorithms.A wrapper-based approach,which tends to give superior performance,and min-max normalization methods were used for feature selection and normalization,respectively.Numerous IDS models were implemented using the full and feature-selected copies of the datasets with and without normalization.The models were evaluated using popular evaluation metrics in IDS modeling,intra-and inter-model comparisons were performed between models and with state-of-the-art works.Random forest(RF)models performed better on NSL-KDD and UNSW-NB15 datasets with accuracies of 99.86%and 96.01%,respectively,whereas artificial neural network(ANN)achieved the best accuracy of 95.43%on the CSE–CIC–IDS2018 dataset.The RF models also achieved an excellent performance compared to recent works.The results show that normalization and feature selection positively affect IDS modeling.Furthermore,while feature selection benefits simpler algorithms(such as RF),normalization is more useful for complex algorithms like ANNs and deep neural networks(DNNs),and algorithms such as Naive Bayes are unsuitable for IDS modeling.The study also found that the UNSW-NB15 and CSE–CIC–IDS2018 datasets are more complex and more suitable for building and evaluating modern-day IDS than the NSL-KDD dataset.Our findings suggest that prioritizing robust algorithms like RF,alongside complex models such as ANN and DNN,can significantly enhance IDS performance.These insights provide valuable guidance for managers to develop more effective security measures by focusing on high detection rates and low false alert rates.
基金supported by the Basic Science Research Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Science and ICT(NRF-2022R1A2C2012243).
文摘In recent decades,brain tumors have emerged as a serious neurological disorder that often leads to death.Hence,Brain Tumor Segmentation(BTS)is significant to enable the visualization,classification,and delineation of tumor regions in Magnetic Resonance Imaging(MRI).However,BTS remains a challenging task because of noise,non-uniform object texture,diverse image content and clustered objects.To address these challenges,a novel model is implemented in this research.The key objective of this research is to improve segmentation accuracy and generalization in BTS by incorporating Switchable Normalization into Faster R-CNN,which effectively captures the fine-grained tumor features to enhance segmentation precision.MRI images are initially acquired from three online datasets:Dataset 1—Brain Tumor Segmentation(BraTS)2018,Dataset 2—BraTS 2019,and Dataset 3—BraTS 2020.Subsequently,the Switchable Normalization-based Faster Regions with Convolutional Neural Networks(SNFRC)model is proposed for improved BTS in MRI images.In the proposed model,Switchable Normalization is integrated into the conventional architecture,enhancing generalization capability and reducing overfitting to unseen image data,which is essential due to the typically limited size of available datasets.The network depth is increased to obtain discriminative semantic features that improve segmentation performance.Specifically,Switchable Normalization captures the diverse feature representations from the brain images.The Faster R-CNN model develops end-to-end training and effective regional proposal generation,with an enhanced training stability using Switchable Normalization,to perform an effective segmentation in MRI images.From the experimental results,the proposed model attains segmentation accuracies of 99.41%,98.12%,and 96.71%on Datasets 1,2,and 3,respectively,outperforming conventional deep learning models used for BTS.
文摘Renormalization group analysis has been proposed to eliminate secular terms in perturbation solutions of differential equations and thus expand the domain of their validity.Here we extend the method to treat periodic orbits or limit cycles.Interesting normal forms could be derived through a generalization of the concept'resonance',which offers nontrivial analytic approximations.Compared with traditional techniques such as multi-scale methods,the current scheme proceeds in a very straightforward and simple way,delivering not only the period and the amplitude but also the transient path to limit cycles.The method is demonstrated with several examples including the Duffing oscillator,van der Pol equation and Lorenz equation.The obtained solutions match well with numerical results and with those derived by traditional analytic methods.
基金supported by the National Natural Science Foundation of China(NSFC,Nos.82373793,82173742)the Science Fund for Distinguished Young Scholars of Shaanxi Province(No.2022JC-54)the Key Research and Development Program of Shaanxi Province(No.2023-YBSF-131).
文摘Tumor vascular normalization has emerged as a promising strategy for synergistic therapy recently.Based on the strategy of“fluorescence turn on-controllable release”,a novel bifunctional candidate was con-structed based on previous developed vascular normalization inducer QDAU5,which could self-assemble to form functional enzyme infrared QDAU5 nanoparticles(FEIRQ NPs).Subsequently,biological evaluation demonstrated that the FEIRQ NPs could induce ferroptosis,endoplasmic reticulum stress,and antigen pre-conditioning and maturation of dendritic cells and CD8^(+)T cells,leading to excellent antitumor efficacy in the absence of cytotoxic drugs.Additionally,FEIRQ NPs show high fluorescence intensity upon expo-sure to theβ-galactosidase(β-Gal)enzyme expressed in ovarian cancer,enabling real-time monitoring of therapeutic effects.Overall,our findings suggest a prospering strategy to early diagnosis and efficient therapy for ovarian cancer without cytotoxicity.
基金supported by the National Natural Science Foundation of China(NSFC,No.82173742)the Science Fund for Distinguished Young Scholars of Shaanxi Province(No.2022JC-54)the Key Research and Development Program of Shaanxi Province(No.2023-YBSF-131).
文摘Remodeling tumor microenvironment(TME)is a very promising and effective strategy to enhance the effects of chemotherapy,photodynamic therapy,and immunotherapy.Normalization of tumor vasculature as well as depletion of glutathione(GSH)can improve the TME.Here,we developed a novel therapeutic nanoparticle functional enzyme ultra QDAU5 nanoparticles(FEUQ Nps)based on a fluorescence-on and releasable strategy by combining a vascular normalization inducer,a GSH depleting agent,and an activated fluorophore.In which the cleavage of disulfide bonds releases active molecules that induce vascular normalization and improve the hypoxic microenvironment.In addition,it may deplete GSH in cancer cells,thus inducing the production of reactive oxygen species(ROS)and lipid peroxide(LPO)and promoting iron toxicity.It may also lead to endoplasmic stress and release of calmodulin,which activates the immune system.Meanwhile,quenched fluorophores are turned on in the presence of galactosidase(GLU)for tumor-specific labeling.In summary,we developed novel therapeutic agent nanoparticles with the function of vascular normalization inducers to achieve specific labeling of hepatocellular carcinoma while exerting efficient antitumor effects in vivo.
文摘Due to differences in the distribution of scores for different trials, the performance of a speaker verification system will be seriously diminished if raw scores are directly used for detection with a unified threshold value. As such, the scores must be normalized. To tackle the shortcomings of score normalization methods, we propose a speaker verification system based on log-likelihood normalization (LLN). Without a priori knowledge, LLN increases the separation between scores of target and non-target speaker models, so as to improve score aliasing of “same-speaker” and “different-speaker” trials corresponding to the same test speech, enabling better discrimination and decision capability. The experiment shows that LLN is an effective method of scoring normalization.
文摘针对说话人确认中,各目标话者模型输出评分分布不一致而导致系统确认阈值设置的困难,本文采取了通过评分规整确定系统最小检测代价函数(DCF)确认阈值的方法。在分析了已有的两种评分规整方法Z norm a l-ization和T norm a lization的基础上,提出了一种结合两者优点的组合规整方法——TZ norm a lization,并据此给出了一种阈值动态修正方法,有效地提高了系统的性能和阈值选取的鲁棒性。对历年的N IST(手机电话语音)评测语料库进行了实验,表明了该方法的有效性。
基金the National Key R&D Program of China(No.2017YFC0906800)the National Natural Science Foundation of China(Nos.81590953,31821002 and 21405020)。
文摘Robust normalization is a prerequisite for reliable metabonomic analysis especially when intervention treatments cause drastic metabolomic changes or when spot urinary samples are employed without knowing the drinking water quantity.With the simulated and real datasets,here,we report a probabilistic quotient normalization method based on the mode-of-quotients(mPQN)which is suitable for metabonomic analysis of both NMR and LC-MS data with little and/or drastic metabolite changes.When applied to metabonomic analysis of both animal plasma samples and human urinary samples,this newly proposed method has clearly shown better robustness than all classical normalization methods especially when drastic changes of some metabolites occur.
基金Supported by the National Basic Research Program of China （2009CB219603, 2010CB226800） the National Natural Science Foundation of China （40874071, 40672104）
文摘During the process of coal prospecting and exploration, different measurement time, different logging instruments and series can lead to systematic errors in well logs. Accordingly, all logging curves need to be normalized in the mining area. By studying well-logging normalization methods, and focusing on the characteristics of the coalfield, the frequency histogram method was used in accordance with the condition of the Guqiao Coal Mine. In this way, the density and sonic velocity at marker bed in the non-key well were made to close to those in the key well, and were eventually equal. Well log normalization was completed when this method was applied to the entire logging curves. The results show that the scales of logging data were unified by normalizing coal logging curves, and the logging data were consistent with wave impedance inversion data. A satisfactory inversion effect was obtained.
基金co-supported by Key Technology Research and Development Program of Jiangsu Province, China (No. BE2011178)the Aviation Industry Innovation Fund (No. AC2011214)
文摘A novel approach is proposed to detect the normal vector to product surface in real time for the robotic precision drilling system in aircraft component assembly, and the auto-normalization algorithm is presented based on the detection system. Firstly, the deviation between the normal vector and the spindle axis is measured by the four laser displacement sensors installed at the head of the multi-function end effector. Then, the robot target attitude is inversely solved according to the auto-normalization algorithm. Finally, adjust the robot to the target attitude via pitch and yaw rotations about the tool center point and the spindle axis is corrected in line with the normal vector simultaneously. To test and verify the auto-normalization algorithm, an experimental platform is established in which the laser tracker is introduced for accurate measurement. The results show that the deviations between the corrected spindle axis and the normal vector are all reduced to less than 0.5°, with the mean value 0.32°. It is demonstrated the detection method and the autonormalization algorithm are feasible and reliable.
基金This paper is supported by the National Natural Science Foundation ofChina (No .40371107) .
文摘In order to evaluate radiometric normalization techniques, two image normalization algorithms for absolute radiometric correction of Landsat imagery were quantitatively compared in this paper, which are the Illumination Correction Model proposed by Markham and Irish and the Illumination and Atmospheric Correction Model developed by the Remote Sensing and GIS Laboratory of the Utah State University. Relative noise, correlation coefficient and slope value were used as the criteria for the evaluation and comparison, which were derived from pseudo-invarlant features identified from multitemporal Landsat image pairs of Xiamen （厦门） and Fuzhou （福州） areas, both located in the eastern Fujian （福建） Province of China. Compared with the unnormalized image, the radiometric differences between the normalized multitemporal images were significantly reduced when the seasons of multitemporal images were different. However, there was no significant difference between the normalized and unnorrealized images with a similar seasonal condition. Furthermore, the correction results of two algorithms are similar when the images are relatively clear with a uniform atmospheric condition. Therefore, the radiometric normalization procedures should be carried out if the multitemporal images have a significant seasonal difference.
基金supported by the National Natural Science Foundation of China(Nos.11472228 and 11202174)the Sichuan Provincial Youth Science and Technology Innovation Team(No.2013TD0004)
文摘This study successfully deals with the inhomogeneous dimension problem of load separation assumption, which is the theoretical basis of the normalization method. According to the dimensionless load separation principle, the normalization method has been improved by intro- ducing a forcible blunting correction. With the improved normalization method, the J-resistance curves of five different metallic materials of CT and SEB specimens are estimated. The forcible blunting correction of initial crack size plays an important role in the J-resistance curve estima- tion, which is closely related to the strain hardening level of material. The higher level of strain hardening leads to a greater difference in JQ determined by different slopes of the blunting line. If the blunting line coefficient recommended by ASTM E1820-11 is used in the improved nor- realization method, it will lead to greater fracture resistance than that processed by the blunting line coefficient recommended by ISO 12135-2002. Therefore, the influence of the blunting line on the determination of JQ must be taken into full account in the fracture toughness assessment of metallic materials.
文摘针对因特网上数字图像的版权保护、认证和完整性等问题,基于DCT变换、image moment normalization和m序列,提出了一种二值水印嵌入算法,实现了二值图像的嵌入和提取。根据m序列的伪随机性和抗干扰性能,使水印具有良好的隐蔽性和稳健性;使用了moment normalization能抵制各种几何攻击。实验表明该算法具有很好的鲁棒性、实用性和可操作性。
基金the National Natural Science Foundation of China,No.81901257(to YXW)the Natural Science Foundation of Jiangsu Province of China,No.BK20180951(to YXW)+1 种基金Postgraduate Research and Practice Innovation Program of Jiangsu Province of China,No.KYCX20_2818(to WL)and Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD,to Key Laboratory of Neuroregeneration of Jiangsu and Ministry of Education).
文摘There is a lack of systematic research on the expression of internal control genes used for gene expression normalization in real-time reverse transcription polymerase chain reaction in spinal cord injury research.In this study,we used rat models of spinal cord hemisection to analyze the expression stability of 13 commonly applied reference genes:Actb,Ankrd27,CypA,Gapdh,Hprt1,Mrpl10,Pgk1,Rictor,Rn18s,Tbp,Ubc,Ubxn11,and Ywhaz.Our results show that the expression of Ankrd27,Ubc,and Tbp were stable after spinal cord injury,while Actb was the most unstable internal control gene.Ankrd27,Ubc,Tbp,and Actb were consequently used to investigate the effects of internal control genes with differing stabilities on the normalization of target gene expression.Target gene expression levels and changes over time were similar when Ankrd27,Ubc,and Tbp were used as internal controls but different when Actb was used as an internal control.We recommend that Ankrd27,Ubc,and Tbp are used as internal control genes for real-time reverse transcription polymerase chain reaction in spinal cord injury research.This study was approved by the Administration Committee of Experimental Animals,Jiangsu Province,China(approval No.20180304-008)on March 4,2018.
基金This research was funded by the National Natural Science Fund of China[grant number 41701415]Science fund project of Wuhan Institute of Technology[grant number K201724]Science and Technology Development Funds Project of Department of Transportation of Hubei Province[grant number 201900001].
文摘Radiometric normalization,as an essential step for multi-source and multi-temporal data processing,has received critical attention.Relative Radiometric Normalization(RRN)method has been primarily used for eliminating the radiometric inconsistency.The radiometric trans-forming relation between the subject image and the reference image is an essential aspect of RRN.Aimed at accurate radiometric transforming relation modeling,the learning-based nonlinear regression method,Support Vector machine Regression(SVR)is used for fitting the complicated radiometric transforming relation for the coarse-resolution data-referenced RRN.To evaluate the effectiveness of the proposed method,a series of experiments are performed,including two synthetic data experiments and one real data experiment.And the proposed method is compared with other methods that use linear regression,Artificial Neural Network(ANN)or Random Forest(RF)for radiometric transforming relation modeling.The results show that the proposed method performs well on fitting the radiometric transforming relation and could enhance the RRN performance.
文摘Background： Expression levels for genes of interest must be normalized with an appropriate reference, or housekeeping gene, to make accurate comparisons of quantitative real-time PCR results. The purpose of this study was to identify the most stable housekeeping genes in porcine articular cartilage subjected to a mechanical injury from a panel of 10 candidate genes. Results： Ten candidate housekeeping genes were evaluated in three different treatment groups of mechanically impacted porcine articular cartilage. The genes evaluated were： beta actin, beta-2-microglobulin, glyceraldehyde-3-phosphate dehydrogenase, hydroxymethylbilane synthase, hypoxanthine phosphoribosyl transferase, peptidylprolyl isomerase A （cyclophilin A）, ribosomal protein L4, succinate dehydrogenase flavoprotein subunit A, TATA box binding protein, and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein--zeta polypeptide The stability of the genes was measured using geNorm, BestKeeper, and NormFinder software. The four most stable genes measured via geNorm were （most to least stable） succinate dehydrogenase flavoprotein, subunit A, peptidylprolyl isomerase A, glyceraldehyde-3-phosphate dehydrogenase, beta actin; the four most stable genes measured via BestKeeper were glyceraldehyde-3-phosphate dehydrogenase, peptidylprolyl isomerase A, beta actin, succinate dehydrogenase flavoprotein, subunit A; and the four most stable genes measured via NormFinder were peptidylprolyl isomerase A, sucdnate dehydrogenase flavoprotein, subunit A, glyceraldehyde-3-phosphate dehydrogenase, beta actin. Conclusions： BestKeeper, geNorm, and NormFinder all generated similar results for the most stable genes in porcine articular cartilage. The use of these appropriate reference genes will facilitate accurate gene expression studies of porcine articular cartilage and suggest appropriate housekeeping genes for articular cartilage studies in other species.
基金National Natural Science Foundation of China(No.59779005)
文摘The hydrodynamic coefficients C-d and C-m are not only dependent on the size of slender cylinder, its location in water, KC number and Re number, but also vary with environmental conditions, i.e., in regular waves or in irregular waves, in pure waves or in wave-current coexisting field. In this paper, the normalization of hydrodynamic coefficients for various environmental conditions is discussed. When a proper definition of KC number and proper characteristic values of irregular waves are used, a unified relationship between C-d, C-m and KC number for regular waves, irregular waves, pure waves and wave-current coexisting field can be obtained.
基金Supported by the Knowledge Innovation Program of Chinese Academy of Sciences(No.KSCX2-EW-G-12B)the Knowledge Innovation Program of the Chinese Academy of Sciences(No.KZCX2-EW-Q213)the National High Technology Research and Development Program of China (863 Program)(No.2012AA10A412)
文摘Quantitative real-time reverse transcription-polymerase chain reaction （qRT-PCR） is widely used in studies of gene expression. In most of these studies, housekeeping genes are used as internal references without validation. To identify appropriate reference genes for qRT-PCR in Pacific abalone Haliotis discus hannai, we examined the transcription stability of six housekeeping genes in abalone tissues in the presence and absence of bacterial infection. For this purpose, abalone were infected with the bacterial pathogen Fibrio anguillarum for 12 h and 48 h. The mRNA levels of the housekeeping genes in five tissues （digestive glands, foot muscle, gill, hemocyte, and mantle） were determined by qRT-PCR. The PCR data was subsequently analyzed with the geNorm and NormFinder algorithms. The results show that in the absence of bacterial infection, elongation factor-l-alpha and beta-actin were the most stably expressed genes in all tissues, and thus are suitable as cross-tissue type normalization factors. However, we did not identify any universal reference genes post infection because the most stable genes varied between tissue types. Furthermore, for most tissues, the optimal reference genes identified by both algorithms at 12 h and 48 h post-infection differed. These results indicate that bacterial infection induced significant changes in the expression of abalone housekeeping genes in a manner that is dependent on tissue type and duration of infection. As a result, different normalization factors must be used for different tissues at different infection points.
基金Project supported by the National Natural Science Foundation of China(Nos.11102113 and81301816)the New Teachers Start Program of Shanghai Jiao Tong University+1 种基金the Chenxing Young Scholars Program B of Shanghai Jiao Tong University(No.13X100010070)the Natural Science Research Foundation of Shanghai Jiao Tong University School of Medicine(No.13XJ10037)
文摘The changes of blood perfusion and oxygen transport in tumors during tumor vascular normalization are studied with 3-dimensional mathematical modeling and numerical simulation. The models of tumor angiogenesis and vascular-disrupting are used to simulate ＂un-normalized＂ and ＂normalized＂ vasculatures. A new model combining tumor hemodynamics and oxygen transport is developed. In this model, the intravasculartransvascular-interstitial flow with red blood cell（RBC） delivery is tightly coupled, and the oxygen resource is produced by heterogeneous distribution of hematocrit from the flow simulation. The results show that both tumor blood perfusion and hematocrit in the vessels increase, and the hypoxia microenvironment in the tumor center is greatly improved during vascular normalization. The total oxygen content inside the tumor tissue increases by about 67%, 51%, and 95% for the three approaches of vascular normalization,respectively. The elevation of oxygen concentration in tumors can improve its metabolic environment, and consequently reduce malignancy of tumor cells. It can also enhance radiation and chemotherapeutics to tumors.
