深度学习在视频对象分割中的应用与展望

视频对象分割是指在给定的一段视频序列的各帧图像中,找出属于特定前景对象的所有像素点位置区域.随着硬件平台计算能力的提升,深度学习受到了越来越多的关注,在视频对象分割领域也取得了一定的进展.本文首先介绍了视频对象分割的主要任务,并总结了该任务所面临的挑战.其次,对开放的视频对象分割常用数据集进行了简要概述,并介绍了通用的性能评估标准.接着,综述了视频对象分割的研究现状,详细地分析了当前的各种方法,并将它们划分为三大类:半监督的方法,即给出视频第一帧图像中感兴趣对象的详细人工真值标注,分割出视频剩余图像中的感兴趣对象;无监督的方法,即不给任何人工标注信息,自动识别并分割出视频中的前景对象;交互式的方法,即在分割过程中,通过人工交互式的参与,结合粗略的人工标注先验信息,进行视频对象分割.第三类方法的条件相当于前两者的折中:相对于第一类方法,它虽然需要人工的参与,但只需要少量的标注工作量;相对于第二类方法,它给视频序列中某些帧的图像适当地添加了一些人工标注信息,从而更具针对性.最后,对深度学习在视频对象分割任务中的应用,进行了总结和展望.     

Automatic Channel Detection Using DNN on 2D Seismic Data

Geologists interpret seismic data to understand subsurface properties and subsequently to locate underground hydrocarbon resources. Channels are among the most important geological features interpreters analyze to locate petroleum reservoirs. However, manual channel picking is both time consuming and tedious. Moreover, similar to any other process dependent on human intervention, manual channel picking is error prone and inconsistent. To address these issues, automatic channel detection is both necessary and important for efficient and accurate seismic interpretation. Modern systems make use of real-time image processing techniques for different tasks. Automatic channel detection is a combination of different mathematical methods in digital image processing that can identify streaks within the images called channels that are important to the oil companies. In this paper, we propose an innovative automatic channel detection algorithm based on machine learning techniques. The new algorithm can identify channels in seismic data/images fully automatically and tremendously increases the efficiency and accuracy of the interpretation process. The algorithm uses deep neural network to train the classifier with both the channel and non-channel patches. We provide a field data example to demonstrate the performance of the new algorithm. The training phase gave a maximum accuracy of 84.6% for the classifier and it performed even better in the testing phase, giving a maximum accuracy of 90%.     

Stock Price Forecasting: An Echo State Network Approach

Forecasting stock prices using deep learning models suffers from problems such as low accuracy, slow convergence, and complex network structures. This study developed an echo state network (ESN) model to mitigate such problems. We compared our ESN with a long short-term memory (LSTM) network by forecasting the stock data of Kweichow Moutai, a leading enterprise in China’s liquor industry. By analyzing data for 120, 240, and 300 days, we generated forecast data for the next 40, 80, and 100 days, respectively, using both ESN and LSTM. In terms of accuracy, ESN had the unique advantage of capturing nonlinear data. Mean absolute error (MAE) was used to present the accuracy results. The MAEs of the data forecast by ESN were 0.024, 0.024, and 0.025, which were, respectively, 0.065, 0.007, and 0.009 less than those of LSTM. In terms of convergence, ESN has a reservoir state-space structure, which makes it perform faster than other models. Root-mean-square error (RMSE) was used to present the convergence time. In our experiment, the RMSEs of ESN were 0.22, 0.27, and 0.26, which were, respectively, 0.08, 0.01, and 0.12 less than those of LSTM. In terms of network structure, ESN consists only of input, reservoir, and output spaces, making it a much simpler model than the others. The proposed ESN was found to be an effective model that, compared to others, converges faster, forecasts more accurately, and builds time-series analyses more easily.     

Semisupervised Encrypted Traffic Identification Based on Auxiliary Classification Generative Adversarial Network

The rapidly increasing popularity of mobile devices has changed the methods with which people access various network services and increased network traffic markedly. Over the past few decades, network traffic identification has been a research hotspot in the field of network management and security monitoring. However, as more network services use encryption technology, network traffic identification faces many challenges. Although classic machine learning methods can solve many problems that cannot be solved by port- and payload-based methods, manually extract features that are frequently updated is time-consuming and labor-intensive. Deep learning has good automatic feature learning capabilities and is an ideal method for network traffic identification, particularly encrypted traffic identification; Existing recognition methods based on deep learning primarily use supervised learning methods and rely on many labeled samples. However, in real scenarios, labeled samples are often difficult to obtain. This paper adjusts the structure of the auxiliary classification generation adversarial network (ACGAN) so that it can use unlabeled samples for training, and use the wasserstein distance instead of the original cross entropy as the loss function to achieve semisupervised learning. Experimental results show that the identification accuracy of ISCX and USTC data sets using the proposed method yields markedly better performance when the number of labeled samples is small compared to that of convolutional neural network (CNN) based classifier.     

众筹项目推荐:面向隐式反馈的深度学习协同过滤

目前,在推荐系统研究中,用户的隐式反馈,以及极度稀疏的数据,已成为影响协同过滤推荐效果的主要问题.针对这一现象,本文提出了深度学习协同过滤算法,先利用卷积神经网络,对用户-项目矩阵的隐层特征进行学习,再结合协同过滤,对用户-项目的交互信息进行建模,并将两种特征融合预测推荐列表.以众筹平台的数据为实验对象,比较模型中各参数对推荐效果的影响,并设计与基线方法的对比实验.实验结果表明:均匀采集负反馈,并在一定卷积层数的网络中,数据稀疏度越高,效果越好;对比基线方法,本文提出的算法在公开数据集(Yahoo!Movie)上取得了最好的推荐结果.本文提出的算法有助于提高众筹平台的融资成功率,同时也丰富了推荐系统的研究体系.     

一种鲁棒性的少样本学习方法

少样本学习是目前机器学习研究领域的一个热点,它能在少量的标记样本中学习到较好的分类模型.但是,在噪声的不确定环境中,传统的少样本学习模型泛化能力弱.针对这一问题,提出一种鲁棒性的少样本学习方法RFSL(Robust Few-Shot Learning).首先,使用核密度估计(Kernel Density Estimation,KDE)和图像滤波(Image Filtering)方法在训练集中加入不同的随机噪声,形成多个不同噪声下的训练集,并分别生成支持集和查询集.其次,利用关系网络的关系模块通过训练集端到端地学习多个基分类器.最后,采用投票的方式对各基分类器的最末Sigmoid层非线性分类结果进行融合.实验结果表明,RFSL模型可促进小样本学习快速收敛,同时,与R-Net以及其他主流少样本学习方法相比,RFSL具有更高的分类准确率,更强的鲁棒性.     

注意力ConvLSTM模型在RUL预测中的应用

预测性维护的应用能够极大地降低企业运维成本,而设备剩余使用寿命(Remaining Useful Life,RUL)预测是预测性维护的关键技术之一.针对传统RUL预测算法难以提取时序数据的潜藏特征以及特征权重分配不合理的问题,本文提出一种基于注意力机制(Attention Mechanism)的卷积长短时记忆(Convolution Long-Short Term Memory,ConvLSTM)预测模型,该模型充分利用LSTM网络处理和预测长期时间序列的优势,并引入注意力机制对产生显著影响的特征因子提高权重,极大地优化了模型的时空特征提取能力.为验证模型预测效果,本文以NASA提供的CMAPSS数据集为对象进行实验,以均方根误差(Root Mean Squared Error,RMSE)和数据集自定义的Score为评价指标,将预测结果与其他RUL预测算法作比对,证明了该模型具有更佳的预测准确性.     

深度学习的图像实例分割方法综述

实例分割是一项具有挑战性的任务,需要同时进行实例级和像素级的预测,在自动驾驶、视频分析、场景理解等方面应用广泛.近年来,基于深度学习的实例分割方法迅速发展,如两阶段检测器Faster R-CNN扩展出的聚焦于网络的精度而非速度的强大实例分割基准Mask R-CNN,一度成为实例分割的标杆.利用高速检测的单阶段检测器延伸出的实例分割算法YOLACT填补了实时实例分割模型的空白,具有较高的研究和应用价值.本文首先对实例分割算法进行了类别划分,然后对一些代表性的算法及其改进算法进行了深入分析,并阐述了相关算法的优缺点,最后对实例分割方法未来的发展进行了展望.     

一种基于图卷积网络的文本多标签学习方法

多标签学习广泛应用于文本分类、标签推荐、主题标注等.最近,基于深度学习技术的多标签学习受到广泛关注,针对如何在多标签学习中有效挖掘并利用高阶标签关系的问题,提出一种基于图卷积网络探究标签高阶关系的模型TMLLGCN.该模型采用GCN的映射函数从数据驱动的标签表示中生成对象分类器挖掘标签高阶关系.首先,采用深度学习方法提取文本特征,然后以数据驱动方式获得基础标签关联表示矩阵,为更好地建模高阶关系及提高模型效果,在基础标签关联表示矩阵上考虑未标记标签集对已知标签集的影响进行标签补全,并以此相关性矩阵指导GCN中标签节点之间的信息传播,最后将提取的文本特征应用到学习高阶标签关系的图卷积网络分类器进行端到端训练,综合标签关联和特征信息作为最终的预测结果.在实际多标签数据集上的实验结果表明,提出的模型能够有效建模标签高阶关系且提升了多标签学习的效果.     

高速网络流量管理系统研究

随着互联网用户的不断增加和用户需求的不断变化,互联网上的应用类型也在发生着日新月异的变化。各种类型的业务流量不断增长,网络的负担不断加重,网络管理的难度也随之增加。如何通过合适的流量管理技术,即网络应用协议的识别与带宽管理技术,来协调网络应用与安全、带宽增长与业务收益、网络扩容与用户体验之间的关系,是网络管理者与运营者亟待解决的重要挑战。文章的研究便着重于此,首先介绍了网络流量管理的基本概念与关键技术,随后对主流的开源流量管理系统和商用流量管理系统进行介绍。最后提出了一种协同式的分布流量管理系统,分布式的部署流量管理子系统通过分析控制中心协同检测与管理流量,可以有效的提高网络流量管理效率,提高未知网络协议的识别率,能够有效检测和防止DDos攻击。