改进Smaller VGGNet的细粒度汽车图像分类算法

为了提高细粒度车辆图像分类的准确度,提出了一种基于改进Smaller VGGNet的细粒度汽车图像分类算法模型.首先,重新调整了神经网络结构,以此提高训练的稳定性和分类的准确率;其次,替换特征的填充方式,以更好地捕捉局部信息,减缓过拟合现象;最后,利用ELU替换原有的ReLU激活函数,以加快模型收敛速度.在公开数据集上,将改进后模型与现有的6种图像分类算法进行对此.结果表明,改进后模型的细粒度汽车图像分类效果显著,且在训练过程表现出更好的稳定性.研究结果可为同类问题解决提供借鉴.

Smaller & Smarter: Score-Driven Network Chaining of Smaller Language Models

With the continuous evolution and expanding applications of Large Language Models (LLMs), there has been a noticeable surge in the size of the emerging models. It is not solely the growth in model size, primarily measured by the number of parameters, but also the subsequent escalation in computational demands, hardware and software prerequisites for training, all culminating in a substantial financial investment as well. In this paper, we present novel techniques like supervision, parallelization, and scoring functions to get better results out of chains of smaller language models, rather than relying solely on scaling up model size. Firstly, we propose an approach to quantify the performance of a Smaller Language Models (SLM) by introducing a corresponding supervisor model that incrementally corrects the encountered errors. Secondly, we propose an approach to utilize two smaller language models (in a network) performing the same task and retrieving the best relevant output from the two, ensuring peak performance for a specific task. Experimental evaluations establish the quantitative accuracy improvements on financial reasoning and arithmetic calculation tasks from utilizing techniques like supervisor models (in a network of model scenario), threshold scoring and parallel processing over a baseline study.

Diagnostic sensitivity of imaging modalities for hepatocellular carcinoma smaller than 2 cm

AIM:To compare the imaging results with histology and to evaluate the diagnostic sensitivity of imaging modalities for hepatocellular carcinoma(HCC)smaller than 2 cm.METHODS:Nodules smaller than 2 cm(n=34)revealed by ultrasonography(US)in 29 patients with liver cirrhosis were analyzed.Histological diagnosis of HCC was performed by ultrasonographic guidance:moderately-differentiated HCC(n=24);well-differentiated HCC(n=10).The patterns disclosed by the four imaging modalities defined the conclusive diagnosis of HCC:(1)contrast-enhanced computed tomography(CECT),hypervascularity in the arterial phase and washout in the equilibrium phase;(2)Sonazoid contrast-enhanced US(CEUS),hypervascularity in the early vascular phase and defect in the Kupffer phase;(3)gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid(Gd-EOBDTPA)-enhanced magnetic resonance imaging(MRI),hypervascularity in the arterial phase and/or defect in the hepatobiliary phase;and(4)CT arterioportal angiography:hypervascularity by CT during arteriography and/ or perfusion defect by CT during arterial portography.RESULTS:Overall,the sensitivity of diagnosing HCC smaller than 2 cm was 52.9%(18/34)(95%CI:35.170.2)by CECT;67.6%(23/34)(95%CI:49.5-82.6)by Sonazoid CEUS;76.5%(26/34)(95%CI:58.8-89.3) by Gd-EOB-DTPA MRI;and 88.2%(30/34)(95%CI: 72.5-96.7)by CT arterioportal angiography.The diagnostic sensitivity of detecting moderately-differentiated HCC by CECT,Sonazoid CEUS,Gd-EOB-DTPA MRI and CT arterioportal angiography was 62.5%(15/24)(95%CI: 40.6-81.2),79.2%(19/24)(95%CI:57.8-92.9),75.0% (18/24)(95%CI:53.3-90.2)and 95.8%(23/24)(95% CI:78.9-99.9),respectively.A significant difference(P< 0.05)was observed between CECT and CT arterioportal angiography in all nodules.There was no difference between Sonazoid CEUS,Gd-EOB-DTPA MRI,and CT arterioportal angiography.The combined sensitivity of Sonazoid CEUS and Gd-EOB-DTPA MRI was 94.1%(32/34).CONCLUSION:Changing the main diagnostic modality for HCC smaller than 2 cm from CT arterioportal angiography to Sonazoid CEUS and Gd-EOB-DTPA MRI is recommended.

Multisatellite observations of smaller mesoscale eddy generation in the Kuroshio Extension

Smaller mesoscale eddies(SMEs)have an important effect on the transmission of ocean temperatures,salinity,energy,and marine biochemical processes.However,traditional altimeters,the dominant sensors used to identify and track eddies,have made it challenging to observe SMEs accurately due to resolution limitations.Eddies drive local upwelling or downwelling,leaving signatures on sea surface temperatures(SSTs)and chlorophyll concentrations(Chls).SST can be observed by spaceborne infrared sensors,and Chl can be measured by ocean color remote sensing.Therefore,multisatellite observations provide an opportunity to obtain information to characterize SMEs.In this paper,an eddy detection algorithm based on SST and Chl images is proposed,which identifies eddies by characterizing the spatial and temporal distribution of SST and Chl data.The algorithm is applied to characterize and analyze SMEs in the Kuroshio Extension.Statistical results on their distribution and seasonal variability are shown,and the formation processes are preliminarily discussed.SMEs generation may be contributed by horizontal strain instability,the interaction of topographic obstacles and currents,and wind stress curl.

Faults and Claims about Thermal Environments in Relation to Energy Saving Measures in Smaller Office Buildings

Building equipment, energy-saving systems, and claims of inappropriate indoor thermal environments were analyzed in relation to the floor area using responses to a questionnaire survey of service managers of 157 buildings built in Osaka, Kyoto and Hyogo prefectures in Kinki area of Japan. Results show the following： （1） In smaller buildings （〈 5,000 m2）, setting temperatures are higher in summer and lower in winter, effects of ＂uncomfortable radiation from windows＂ are greater, energy-saving systems decrease indoor thermal comfort, but claims of ＂hot＂ and ＂cold＂ are fewer; （2） Claims of ＂hot＂ and ＂cold＂ are unrelated to the setting temperature and whether the air-conditioning control system is central or local; （3） The adoption rates of mitigation of dress codes （＂COOL-BIZ＂ and ＂WARM-BIZ＂） are higher than those of temperature mitigation of air conditioning.

Spending Power on the rise in China’s Smaller Cities

China’s big spenders now far-flung by Yu Nan WIDELY recognized as one of the most beautiful cities in China,the coastal port of Dalian is a perfect example of a lowertier city becoming a growth hub for multinationals, owing to stronger purchasing power. That trend inspired Boston Consulting Group(BCG) to release a report,advising businessmen to look beyond the familiar mega-cities to the fast-growing smaller

Significant“smaller is softer”in amorphous silicon via irradiation-me diate d surface modification

“Smaller is softer”is a reverse size dependence of strength,defying the“smaller is stronger”tenet.It usually results from surface-mediated displacive or diffusive deformation and is mainly found in some ultra-small-scale(below tens of nanometers)metallic materials.Here,making use of the surface modifi-cation via ion beam irradiation,we bring the“smaller is softer”into being in a covalently-bonded,hard,and brittle material-amorphous Si(a-Si)at a much larger size regime(<∼500 nm).It is manifested as the transition from the quasi-brittle failure to the homogeneous plastic deformation as well as the de-creasing yield stress with sample volume reduction at the submicron-scale regime.An analytical model of hard core/superplastic shell has been proposed to explain the artificially-controllable size-dependent softening.This surface engineering pathway via ion irradiation is not only of particular interest to tai-lor the strength and deformation behaviors in small-sized a-Si or other covalently-bonded amorphous solids but also of practical relevance to the utility of a-Si in microelectronics and microelectromechanical systems.

木豆小叶突变体的表型鉴定及遗传分析

叶片是植物体与周围环境之间进行物质交流与能量转化的主要器官,对植株的生长发育起着关键的作用。本研究对前期航空诱变获得的木豆(Cajanuscajan)小叶突变体进行了表型鉴定。结果表明:其叶宽、叶长、叶面积与野生型相比均显著变小(P<0.01),而叶片长宽比显著变大(P<0.01)。通过对小叶突变体和野生型‘ICPL81-3’杂交构建的F2群体进行表型观察发现,不同单株的叶宽、叶长、长宽比和叶面积在F2群体中均有较大变异,且显著相关,其中叶宽和叶面积呈现明显双峰分布,叶宽表型符合3∶1分离(X^(2)<X^(2)_(0.05,1)=3.84),表明小叶突变性状受隐性单基因控制。利用SSR标记和InDel标记结合混池分离群体分析法(Bulksegregation analysis,BSA)进行小叶基因初定位,获得在两个基因池内具有明显差异的1个InDel标记PA094,经基因池各单株检测和F2群体单株检验,证实PA094和小叶突变性状具有稳定的连锁关系。研究结果为后续开展木豆小叶基因的精细定位及其基因克隆提供了重要基础。

小减压量下制动力差异对重载列车纵向力影响研究

为研究制动力差异对重载列车纵向力的影响,分析了车辆制动力影响因素及车辆间制动力差异的原因。以C 80B型车辆为例,分析了各影响因素的波动范围,仿真计算了不同大小制动力在不同分布工况下的车钩力。结果表明:与制动力均布且均处于平均值为对比,车辆制动力在差异范围内随机分布对列车纵向力影响小;列车以车辆制动力差异分为n部(n≥2),n越小,列车纵向力越大,制动力差异越大,列车纵向力越大。

Brain Tumor Segmentation through Level Based Learning Model

Brain tumors are potentially fatal presence of cancer cells over a human brain,and they need to be segmented for accurate and reliable planning of diag-nosis.Segmentation process must be carried out in different regions based on which the stages of cancer can be accurately derived.Glioma patients exhibit a different level of challenge in terms of cancer or tumors detection as the Magnetic Resonance Imaging(MRI)images possess varying sizes,shapes,positions,and modalities.The scanner used for sensing the location of tumors cells will be sub-jected to additional protocols and measures for accuracy,in turn,increasing the time and affecting the performance of the entire model.In this view,Convolutional Neural Networks deliver suitable models for efficient segmentation and thus delivered promising results.The previous strategies and models failed to adhere to diversity of sizes and shapes,proving to be a well-established solution for detecting tumors of bigger size.Tumors tend to be smaller in size and shape during their premature stages and they can easily evade the algorithms of Convolutional Neural Network(CNN).This proposal intends to furnish a detailed model for sensing early stages of cancer and hence perform segmentation irrespective of the current size and shape of tumors.The size of networks and layers will lead to a significant weightage when multiple kernel sizes are involved,especially in multi-resolution environments.On the other hand,the proposed model is designed with a novel approach including a dilated convolution and level-based learning strat-egy.When the convolution process is dilated,the process of feature extraction deals with multiscale objective and level-based learning eliminates the shortcoming of previous models,thereby enhancing the quality of smaller tumors cells and shapes.The level-based learning approach also encapsulates the feature recon-struction processes which highlights the sensing of small-scale tumors growth.Inclusively,segmenting the images is performed with better accuracy and hence detection becomes better when compared to that of hierarchical approaches.

钢管格构立柱桩在盖挖法高岩面深基坑中的应用研究

针对如何解决高岩面深基坑传统格构立柱桩成孔工效低、垂直度控制低下,影响盖板及基坑支护结构的安全性问题,依据项目实际情况,通过钢管格构立柱桩在盖挖法高岩面深基坑中的应用研究,提出了“大桩变小桩、角钢改钢管”的钢管格构立柱桩替代方法,大大提高了成孔工效、保障格构立柱桩垂直度,提高盖板及基坑支护结构的安全性,有效地解决了传统格构柱在碰到硬度较高的片麻岩、花岗岩等岩层则难于成孔的施工难题。

EH4在危机矿山隐伏金矿体定位预测中的应用研究

EH4连续电导率成像仪是快速高效的第四代数字化大地电磁探测仪器,基本配置(10Hz～100kHz)能测量地表向下1000m深度范围内地质体的连续视电导率值,我们对25个不同类型金属矿床进行了隐伏矿定位预测,取得了显著的效果.2D模式图能清晰地反映不同电性体的精细电阻率结构,经钻孔验证后显示:(1)火山晚期热液型金矿床,很好地反映了火山机构控矿的漏斗状形态,体现了矿体的垂向延深大于水平延长的成矿规律,矿化异常的电阻率较低(1～150Ω·m);(2)剪切带型金矿床,明显区分了矿化蚀变构造带和围岩,确认矿化异常在围岩中呈脉状分布,电阻率小于260Ω·m;(3)隐爆角砾岩型金矿床,清晰地反映了隐爆角砾岩体控矿所形成的对称矿(化)体,区分了隐爆角砾岩体和矿化蚀变带,矿化异常的电阻率较高(1000～2000Ω·m).

茶树不同轻修剪模式对假眼小绿叶蝉种群及茶叶产量、品质的影响

研究了夏剪、冬剪两种轻修剪模式对茶假眼小绿叶蝉种群及茶叶产量、品质的影响。结果表明,1~4月,叶蝉数量少,对茶树为害极轻,虫口密度以夏剪稍高于冬剪;在春茶结束后即实施夏剪,6~10月间其虫口密度明显低于冬剪,表明在叶蝉为害高峰期之前,采取轻修剪措施可有效控制虫口密度,减少虫害损失。不同轻修剪模式对茶叶产量影响较明显,夏剪的春、夏茶产量分别比冬剪高21.3%~36.5%、6.61%~26.3%,对茶叶品质也有明显影响。

砒砂岩区小流域沟冻融风化侵蚀模型研究

文章通过砒砂岩区西召沟小流域冻融风化侵蚀实测资料,分析了冻融风化侵蚀与沟道各变量之间的关系。确定了该流域相同种类支沟上、中、下游的冻融风化坡面侵蚀模数、单位沟沿线沟角线长度和侵蚀量值,不同种类支沟冻融风化坡面蚀模数及单位长度沟沿线沟角线长度和的侵蚀量的比例关系,及上、中、下游冻融风化侵蚀侵蚀量的比例关系。据此建立了小流域单位长度沟沿线沟角线冻融风化侵蚀量比例模型和冻融风化侵蚀模数比例模型。根据西召沟小流域上游20条沟谷的资料,利用两种模型对该流域的冻融风化侵蚀量进行了模拟计算,两种模型的模拟计算结果分别为23757 9t和29165 1t,相差18 5%。

黔南紫云四大寨长兴期小有孔虫

首次报道了产自贵州紫云四大寨剖面长兴期的有孔虫动物群 ,共计有 9属 2 2种。

桂西南柳桥地区二叠纪末期浅水相小有孔虫动物群

本文研究了广西扶绥县柳桥地区二叠纪末期的浅水相小有孔虫动物群。该动物群赋存于硅质岩相地层中的灰岩夹层里。动物群丰度和分异度均较高,共发现12属25种。其中包括长兴期的常见分子如Glomospira,No-dosaria,Colaniella等。此有孔虫组合与Colaniella组合大致相当,可以与类Paleofusulina带和Gallowayiella带相对比。

小尺寸材料弹性常数超声测量系统的研制及其应用

提出一种小尺寸材料弹性常数测量的新方法———利用线聚焦超声传感器和超声浸水技术同时测定纵波和漏表面波的波速,求得材料的弹性常数。基于PXI-8186嵌入式控制器构建了小尺寸材料弹性常数测控系统,利用LabVIEW软件平台开发了小尺寸材料弹性常数自动测量软件。实验结果表明,该方法可以实现小尺寸材料弹性常数的测量,测量结果具有很好的重复性。

浙江省灰飞虱对吡虫啉、锐劲特和毒死蜱的抗药性监测

为明确浙江省灰飞虱抗药性的发展趋势，采集了浙江省不同地区的灰飞虱，监测其对吡虫啉、锐劲特和毒死蜱的抗药性水平。结果表明，浙江省内灰飞虱对吡虫啉已经产生一定抗性，余姚地区最高达到6倍；对于锐劲特和毒死蜱抗性水平较低，其中锐劲特的抗性倍数为1．8～2．8，毒死蜱为1．5～2．4。

小富勒烯研究进展

对近年来C20，C24，C28，C36，C40，C44和C50等小富勒烯及其内嵌、加成和取代衍生物的实验和理论研究进展进行了回顾，并对目前理论研究的热点C2n（n=26～29）进行了讨论．

基于时频分析的电磁轨道发射电感梯度研究

为研究电磁轨道发射试验过程的导轨电感梯度,本文首先利用Matlab时频工具箱对发射试验的电流信号进行处理,得到了发射过程中电流的时频曲线;同时通过Ansoft建模仿真得到了不同频率下的导轨电感梯度值;进而得到实际发射过程中对应于不同时刻的导轨电感梯度值。结果表明在整个发射过程中,工作电流频率是在逐渐减小的,而导轨的电感梯度是在非线性增大;对比由此电感梯度与常值电感梯度(0.40μH/m)引起的推力表明:在发射的初期,二者大小非常接近,但在发射的后期(1.0ms以后)推力差异增大,常值时的推力值较前者偏小约4%。