Fully Distributed Learning for Deep Random Vector Functional-Link Networks

In the contemporary era, the proliferation of information technology has led to an unprecedented surge in data generation, with this data being dispersed across a multitude of mobile devices. Facing these situations and the training of deep learning model that needs great computing power support, the distributed algorithm that can carry out multi-party joint modeling has attracted everyone’s attention. The distributed training mode relieves the huge pressure of centralized model on computer computing power and communication. However, most distributed algorithms currently work in a master-slave mode, often including a central server for coordination, which to some extent will cause communication pressure, data leakage, privacy violations and other issues. To solve these problems, a decentralized fully distributed algorithm based on deep random weight neural network is proposed. The algorithm decomposes the original objective function into several sub-problems under consistency constraints, combines the decentralized average consensus (DAC) and alternating direction method of multipliers (ADMM), and achieves the goal of joint modeling and training through local calculation and communication of each node. Finally, we compare the proposed decentralized algorithm with several centralized deep neural networks with random weights, and experimental results demonstrate the effectiveness of the proposed algorithm.

Spectrum Characteristics and Transfer Function of the Hydrograph of the Deep Aqueous System

The fluctuation of most of the hydrograph in the deep aqueous system records the fluid pulsation in lithosphere and variation of the earth's crust. Many observations have verified that groundwater is an ideal information carrier of the crust. In this paper, the series of input (precipitation, air pressure, Earth tide etc.) and output (water level, artesian flow) of the deep aqueous system are studied by using the spectrum analysis and system theory. The application concepts of transfer function and the spectral structure of the hydrograph enrich the knowledge of the deep aqueous system. Two typical spectral structures of the hydrograph of the deep aqueous system are obtained by comparing with many water-bearing systems of the Jizhong depression. One is from well Ma-17 and the other is from the well Xinze-5. Finally, the physical models of forming the spectrum of the hydrograph are constructed on the basis of the spectrum research on the deep aqueous system.

Functional outcome of tibial fracture with acute compartment syndrome and correlation to deep posterior compartment pressure

AIM To measure single baseline deep posterior compartment pressure in tibial fracture complicated by acute compartment syndrome(ACS) and to correlate it with functional outcome.METHODS Thirty-two tibial fractures with ACS were evaluated clinically and the deep posterior compartment pressure was measured. Urgent fasciotomy was needed in 30 patients. Definite surgical fixation was performed either primarily or once fasciotomy wound was healthy. The patients were followed up at 3 mo, 6 mo and one year. At one year, the functional outcome [lower extremity functional scale(LEFS)] and complications were assessed.RESULTS Three limbs were amputated. In remaining 29 patients, the average times for clinical and radiological union were 25.2 ± 10.9 wk(10 to 54 wk) and 23.8 ± 9.2 wk(12 to 52 wk) respectively. Nine patients had delayed union and 2 had nonunion who needed bone grafting to augment healing. Most common complaint at follow up was ankle stiffness(76%) that caused difficulty in walking,running and squatting. Of 21 patients who had paralysis at diagnosis, 13(62%) did not recover and additional five patients developed paralysis at follow-up. On LEFS evaluation, there were 14 patients(48.3%) with severe disability, 10 patients(34.5%) with moderate disability and 5 patients(17.2%) with minimal disability. The mean pressures in patients with minimal disability, moderate disability and severe disability were 37.8, 48.4 and 58.79 mmH g respectively(P < 0.001).CONCLUSION ACS in tibial fractures causes severe functional disability in majority of patients. These patients are prone for delayed union and nonunion; however, long term disability is mainly because of severe soft tissue contracture. Intracompartmental pressure(ICP) correlates with functional disability; patients with relatively high ICP are prone for poor functional outcome.

短周期密集台阵深部地壳结构探测研究进展

由于穿透能力强,天然地震接收函数成为壳幔结构探测中最为广泛使用的方法.随着人们对地球内部结构和动力学过程认识程度的提高,台间距相对较大的宽频带台阵已无法满足壳幔结构高分辨率探测的需求.短周期密集台阵采用频率较高的便携式数字地震仪,通过百米级台间距的密集观测,可在短时间内(1~2个月)获得大量地震数据.其优势主要表现在三个方面:(1)地壳内射线交叉覆盖好,有利于提高分辨率;(2)射线密集分布,通过相干叠加压制噪声,可实现高频接收函数的成像;(3)观测时间短,效率高.因此,短短几年内,短周期密集台阵已经成为地壳深部结构探测的常规手段之一.本文主要介绍短周期密集台阵深部地壳结构探测的由来,以及通过几个典型的探测实例,展示探测效果及其在不同构造域的应用.

利用函数拟合对DeepFM算法的改进研究

DeepFM模型是基于FM模型与Wide&Deep模型的改进,该推荐算法主要基于深度学习通过已知特征来预测用户点击某一按钮的概率。但随着电子商务的发展,不仅需要通过横向特征预测用户点击某一按钮的概率,还要纵向考虑该按钮在不同时间段的点击概率变化。文中对DeepFM进行了改进,引用了拟合函数的方法,通过各个拟合的函数计算出点击概率变化的函数图像,延展该图像得知该按钮被点击概率随着第三维坐标时间值的变化,从而实现了各种用户在不同时间对于不同商品需求的预测。该算法运用了相对平滑的函数曲线来拟合模型计算的结果,提高了模型的精确度。

基于注意力循环神经网络的联合深度推荐模型

为了向用户推荐符合兴趣偏好的项目,设计一种基于注意力循环神经网络的联合深度推荐模型。将双层注意力机制设置于网络中,该模型由五个部分构成,在输入层中生成联合深度推荐模型的输入矩阵,通过序列编码层对项目评论文本语义展开正向和反向编码,获得隐藏状态输出,并将其输入双层注意力机制中,提取项目特征,利用全连接层提取用户偏好特征。在预测层中建立项目与用户的交互模型,获得项目评分,为用户推荐高评分的项目。为了提高模型精度,加权融合MSE损失函数、CE损失函数和RK损失函数建立组合损失函数,对深度联合训练模型展开训练,提高模型的推荐性能。仿真结果表明,所提方法具有良好的推荐效果,能够适应不断变化的市场需求和用户行为。

基于Deep-IndRNN的DGA域名检测方法

恶意服务常利用域名生成算法(DGA)逃避域名检测,针对DGA域名隐蔽性强、现有检测方法检测速度较慢、实用性不强等问题,采用深度学习技术,提出了一种基于Deep-IndRNN的DGA域名检测方法。方法运用词袋模型(BoW)将域名向量化,然后通过Deep-IndRNN提取域名字符间特征,并使用Sigmoid函数对域名分类检测。其主要特点在于:通过将Deep-IndRNN的多序列输入拼接为单向量输入,以单步处理代替循环处理,同时结合Deep-IndRNN能保存更长时间记忆的特点,可有效释放深度学习时占用的GPU、CPU等系统资源,且在保证高准确率和精确度的前提下提高训练、检测速度。实验结果表明,基于Deep-IndRNN的DGA域名检测方法在检测任务中具有较高的准确率和精确度,相比于DNN、CNN、LSTM、BiLSTM、CNN-LSTM-Concat等同类检测方法,能显著提高训练、检测速度,是有效可行的。

Prediction of Eight Earings in Deep Drawing of 5754O Aluminum Alloy Sheet

Earings appear easily during deep drawing of cylindrical parts owing to the anisotropic properties of materials.However,current methods cannot fully utilize the mechanical properties of material,and the number of earings obtained differ with the simulation methods.In order to predict the eight-earing problem in the cylindrical deep drawing of 5754O aluminum alloy sheet,a new method of combining the yield stress and anisotropy index(r-value)to solve the parameters of the Hil 148 yield function is proposed.The general formula for the yield stress and r-value in any direction is presented.Taking a 5754O aluminum alloy sheet as an example in this study,the deformation area in deep drawing is divided into several equal sectorial regions based on the anisotropy.The parameters of the Hill48 yield function are solved based on the yield stress and r-value simultaneously for the corresponding deformation area.Finite element simulations of deep drawing based on new and existing methods are carried out for comparison with experimental results.This study provides a convenient and reliable way to predict the formation of eight earings in the deep drawing process,which is expected to be useful in industrial applications.The results of this study lay the foundation for the optimization of the cylindrical deep drawing process,including the optimization of the blank shape to eliminate earing defects on the final product,which is of great importance in the actual production process.

Characterization of deep ground geothermal field in Jiahe Coal Mine

Research into the characteristics of geothermal fields is important for the control of heat damage in mines. Based on measured geothermal data of boreholes from 200 m to 1200 m in a Jiahe Coal Mine, we demonstrate non-linear but increasing relations of both geo-temperatures and geothermal gradients with increases depth. Numerically, we fitted the relationship between geo-temperatures and depth, a first-order exponential decay curve, formulated as: T(h) = 4.975 + 23.08 exp(h/1736.1).

基于深度学习的配电网开关柜电晕放电检测设计

在配电网运行过程中,电晕放电量过大导致间隙击穿,开关柜输出电量信号出现过度电离,降低了运行稳定性。为此,提出基于深度学习的配电网开关柜电晕放电检测方法。利用与电晕信号相关的深度学习识别模型,定义单调非递增函数,并以此为基础,计算信号取样值,实现深度学习下的配电网开关柜电晕信号取样。确定电晕信号串表达式,根据开关柜脉冲求解结果,完成实时放电检测。实验结果表明,所提方法可将电晕放电量最大值控制在100 pC以下,能够避免开关柜输出电量信号出现过度电离的情况。

Mathematical Reinforcement to the Minibatch of Deep Learning

We elucidate a practical method in Deep Learning called the minibatch which is very useful to avoid local minima. The mathematical structure of this method is, however, a bit obscure. We emphasize that a certain condition, which is not explicitly stated in ordinary expositions, is essential for the minibatch method. We present a comprehensive description Deep Learning for non-experts with the mathematical reinforcement.

A Novel Approach to Heart Failure Prediction and Classification through Advanced Deep Learning Model

In this study, the author will investigate and utilize advanced machine learning models related to two different methodologies to determine the best and most effective way to predict individuals with heart failure and cardiovascular diseases. The first methodology involves a list of classification machine learning algorithms, and the second methodology involves the use of a deep learning algorithm known as MLP or Multilayer Perceptrons. Globally, hospitals are dealing with cases related to cardiovascular diseases and heart failure as they are major causes of death, not only for overweight individuals but also for those who do not adopt a healthy diet and lifestyle. Often, heart failures and cardiovascular diseases can be caused by many factors, including cardiomyopathy, high blood pressure, coronary heart disease, and heart inflammation [1]. Other factors, such as irregular shocks or stress, can also contribute to heart failure or a heart attack. While these events cannot be predicted, continuous data from patients’ health can help doctors predict heart failure. Therefore, this data-driven research utilizes advanced machine learning and deep learning techniques to better analyze and manipulate the data, providing doctors with informative decision-making tools regarding a person’s likelihood of experiencing heart failure. In this paper, the author employed advanced data preprocessing and cleaning techniques. Additionally, the dataset underwent testing using two different methodologies to determine the most effective machine-learning technique for producing optimal predictions. The first methodology involved employing a list of supervised classification machine learning algorithms, including Naïve Bayes (NB), KNN, logistic regression, and the SVM algorithm. The second methodology utilized a deep learning (DL) algorithm known as Multilayer Perceptrons (MLPs). This algorithm provided the author with the flexibility to experiment with different layer sizes and activation functions, such as ReLU, logistic (sigmoid), and Tanh. Both methodologies produced optimal models with high-level accuracy rates. The first methodology involves a list of supervised machine learning algorithms, including KNN, SVM, Adaboost, Logistic Regression, Naive Bayes, and Decision Tree algorithms. They achieved accuracy rates of 86%, 89%, 89%, 81%, 79%, and 99%, respectively. The author clearly explained that Decision Tree algorithm is not suitable for the dataset at hand due to overfitting issues. Therefore, it was discarded as an optimal model to be used. However, the latter methodology (Neural Network) demonstrated the most stable and optimal accuracy, achieving over 87% accuracy while adapting well to real-life situations and requiring low computing power overall. A performance assessment and evaluation were carried out based on a confusion matrix report to demonstrate feasibility and performance. The author concluded that the performance of the model in real-life situations can advance not only the medical field of science but also mathematical concepts. Additionally, the advanced preprocessing approach behind the model can provide value to the Data Science community. The model can be further developed by employing various optimization techniques to handle even larger datasets related to heart failures. Furthermore, different neural network algorithms can be tested to explore alternative approaches and yield different results.

Probing Nucleon Structure in Deep Inelastic Scattering

The comparison between the muon and the neutrino as probes of the nucleon structure is presented. The prediction of the structure functions, quark distributions, leptonic currents, and cross section led us to obtain some of the features of the electro-weak interactions in the deep inelastic scattering. A perturbation technique is used to evaluate the leptonic current that is assumed to be a complex quantity. The imaginary part of which represents the rate of absorption. On the other hand, the quarks wave functions forming the nucleon are extracted from experimental data for neutrino-nucleon and muon-nucleon collisions. A numerical technique is applied to analyze the data of the experiments CERN-NA-2 and CERN-WA25, to evaluate the quark functions and hence to calculate the hadronic current. It is found that the quark distribution functions predicted by the muon as a probe is slightly shifted up compared with that of the neutrino. Finally, the differential cross section is calculated in terms of leptonic and hadronic currents.

Brain Encoding and Decoding in fMRI with Bidirectional Deep Generative Models

Brain encoding and decoding via functional magnetic resonance imaging(fMRI)are two important aspects of visual perception neuroscience.Although previous researchers have made significant advances in brain encoding and decoding models,existing methods still require improvement using advanced machine learning techniques.For example,traditional methods usually build the encoding and decoding models separately,and are prone to overfitting on a small dataset.In fact,effectively unifying the encoding and decoding procedures may allow for more accurate predictions.In this paper,we first review the existing encoding and decoding methods and discuss the potential advantages of a“bidirectional”modeling strategy.Next,we show that there are correspondences between deep neural networks and human visual streams in terms of the architecture and computational rules.Furthermore,deep generative models(e.g.,variational autoencoders(VAEs)and generative adversarial networks(GANs))have produced promising results in studies on brain encoding and decoding.Finally,we propose that the dual learning method,which was originally designed for machine translation tasks,could help to improve the performance of encoding and decoding models by leveraging large-scale unpaired data.

A Deep Web Query Interfaces Classification Method Based on RBF Neural Network

This paper proposes a new approach for classification for query interfaces of Deep Web, which extracts features from the form＇s text data on the query interfaces, assisted with the synonym library, and uses radial basic function neural network （RBFNN） algorithm to classify the query interfaces. The applied RBFNN is a kind of effective feed-forward artificial neural network, which has a simple networking structure but features with strength of excellent nonlinear approximation, fast convergence and global convergence. A TEL_8 query interfaces＇ data set from UIUC on-line database is used in our experiments, which consists of 477 query interfaces in 8 typical domains. Experimental results proved that the proposed approach can efficiently classify the query interfaces with an accuracy of 95.67%.

FLANGE EARING AND ITS CONTROL ON DEEP-DRAWING OF ANISOTROPY CIRCULAR SHEETS

The Hill's quadric anisotropy yield function and the Barlat-Lian anisotropy yield func- tion describing well anisotropy sheet metal with stronger texture are introduced into a quadric-flow cor- ner constitutive theory of elastic-plastic finite deformation suitable for deformation localization analy- sis.And then,the elastic-plastic large deformation finite element formulation based on the virtual power principle and the discrete Kirchhoff shell element model including the yield functions and the constitutive theory are established.The focus of the present research is on the numerical simulation of the flange earing of the deep-drawing of anisotropy circular sheets,based on the investigated results, the.schemes for controlling the flange earing are proposed.

手术患者术后下肢深静脉血栓形成的术中预防方案分析

目的分析预防手术患者术后下肢深静脉血栓形成(LDVT)的术中应用方案。方法收集2022年3月至2023年2月于保定市第二中心医院收治的256例手术患者的临床资料,将2022年9月前未应用术中改良方案的患者作为常规组(n=119),将2022年9月开始应用术中改良方案的患者作为改良组(n=137)。比较两组患者的手术指标(手术时间、术中出血量、术后卧床时间、住院时间和配合度评分)、血流动力学指标(心率、平均动脉压)、手术前后股静脉血流指标(峰值血流速度、平均血流速度和血流量)、手术前后凝血功能指标[凝血酶原时间(PT)、活化部分凝血活酶时间(APTT)、纤维蛋白原(FIB)、凝血酶时间(TT)和D-二聚体(D-D)]及术后LDVT的发生情况。结果改良组患者的手术时间、术后卧床时间、住院时间均明显短于常规组患者,术中出血量明显少于常规组患者,配合度评分明显高于常规组患者,差异均有统计学意义(P﹤0.01)。入室后,两组患者的心率、平均动脉压比较,差异均无统计学意义(P﹥0.05);术中10 min、术毕即刻,改良组患者的心率、平均动脉压均低于常规组患者,差异均有统计学意义(P﹤0.05)。术后,改良组患者股静脉的峰值血流速度、平均血流速度、血流量均高于常规组患者,差异均有统计学意义(P﹤0.01)。术后,改良组患者的PT、APTT、TT均长于常规组患者,D-D、FIB水平均低于常规组患者,LDVT的发生率低于常规组患者,差异有统计学意义(P﹤0.05)。结论手术室人员于术中应用改良方案能够改善患者术中血流动力学,提高手术配合度,降低术中出血量,从而减轻术后血液高凝状态,促进术后下肢静脉血液回流,降低术后LDVT的发生风险。

融合累积分布函数和通道注意力机制的DeepLabV3+图像分割算法

为了解决DeepLabV3+在语义分割时未充分利用主干的低级特征,以及大倍数上采样造成有效特征缺失的问题,提出一种累积分布通道注意力DeepLabV3+(CDCA-DLV3+)模型。首先,基于累积分布函数和通道注意力提出了累积分布通道注意力(CDCA);然后,利用CDCA获取主干部分有效的低级特征;最后,采用特征金字塔网络(FPN)进行特征融合和逐步上采样,从而避免大倍数上采样所造成的特征损失。CDCA-DLV3+模型在Pascal VOC2012验证集与Cityscapes数据集上的平均交并比(mIoU)分别为80.09%和80.11%,相较于DeepLabV3+模型分别提升1.24和1.02个百分点。实验结果表明,所提模型分割结果更加精确。

行人搜索算法综述

随着深度学习技术的快速发展,行人搜索算法的研究得到大量学者的关注。行人搜索是在行人检测和行人重识别任务的基础上在图像中寻找特定目标行人。该文对近年来行人搜索任务相关研究进展进行了详细梳理。按照模型网络结构和损失函数两方面对现有方法进行分析和总结。依据卷积神经网络和Transformer两类不同的技术路线,重点阐述各自代表性方法的主要研究工作;并按照传统损失函数、OIM损失函数及混合损失函数对行人搜索采用的训练损失函数进行详细总结。此外,总结了行人搜索任务领域常用的公开数据集,比较和分析了主要算法在相应数据集上的性能表现。最后总结了行人搜索任务的未来研究方向。

基于改进YOLOv3-SPP算法的道路车辆检测

针对在城市道路场景下视觉检测车辆时,车辆密集和远处车辆呈现小尺度,导致出现检测精度低或者漏检的问题,提出了一种基于改进的YOLOv3-SPP算法,对激活函数进行优化,以DIOU-NMS Loss作为边界框损失函数,增强网络的表达能力。为提高所提算法对小目标和遮挡目标的特征提取能力,引入空洞卷积模块,增大目标的感受野。实验结果表明,所提算法在检测车辆目标时m AP提高了1.79%,也有效减少了在检测紧密车辆目标时出现的漏检现象。