本文選題：深度學(xué)習(xí) + 卷積神經(jīng)網(wǎng)絡(luò)�。� 參考：《中國(guó)民航大學(xué)》2017年碩士論文

【摘要】：卷積神經(jīng)網(wǎng)絡(luò)作為深度學(xué)習(xí)的一個(gè)分支,已經(jīng)在多個(gè)領(lǐng)域取得了巨大成功,其中神經(jīng)網(wǎng)絡(luò)的深度是其取得成功的關(guān)鍵。然而,越深的神經(jīng)網(wǎng)絡(luò)訓(xùn)練起來(lái)就越困難,因此,論文將主要研究如何設(shè)計(jì)并訓(xùn)練一個(gè)深度卷積神經(jīng)網(wǎng)絡(luò)模型。論文主要工作如下:首先,論文結(jié)合卷積神經(jīng)網(wǎng)絡(luò)的特點(diǎn),提出了一種單通道卷積神經(jīng)網(wǎng)絡(luò)(Single-channel Convolution Neural Networks,SCNN)模型,詳細(xì)介紹了該模型的實(shí)現(xiàn)方式和訓(xùn)練流程。為了降低模型過(guò)擬合的風(fēng)險(xiǎn),將Dropout算法引入SCNN,提高SCNN的泛化能力。為了加快模型訓(xùn)練速度,將批歸一化(Batch Normalization,BN)算法引入SCNN模型,對(duì)網(wǎng)絡(luò)所有卷積層的激活值進(jìn)行批歸一化處理。然后,為解決深度卷積神經(jīng)網(wǎng)絡(luò)由于梯度消失而導(dǎo)致訓(xùn)練困難的問(wèn)題,提出一種快速、高效的雙通道卷積神經(jīng)網(wǎng)絡(luò)(Dual-Channel Convolution Neural Networks,DCNN)模型,該模型由兩種通道構(gòu)成:直通通道和卷積通道。直通通道負(fù)責(zé)保障深度網(wǎng)絡(luò)的暢通性;卷積通道負(fù)責(zé)深度網(wǎng)絡(luò)的學(xué)習(xí)。考慮到深層網(wǎng)絡(luò)在訓(xùn)練時(shí)容易出現(xiàn)性能不穩(wěn)定的問(wèn)題,在卷積通道上引入卷積衰減因子,對(duì)其響應(yīng)數(shù)據(jù)進(jìn)行約束。為了保證各通道數(shù)據(jù)維度的一致性,設(shè)計(jì)了一種雙池化層對(duì)同一特征圖進(jìn)行降采樣。在CIFAR-10、CIFAR-100和MNIST 3個(gè)標(biāo)準(zhǔn)圖像識(shí)別數(shù)據(jù)集上,DCNN分別取得了94.53%,73.40%和99.74%的分類(lèi)準(zhǔn)確率,無(wú)論是神經(jīng)網(wǎng)絡(luò)的可訓(xùn)練深度、穩(wěn)定性和分類(lèi)精度,DCNN都明顯優(yōu)于現(xiàn)有的大多數(shù)深度卷積神經(jīng)網(wǎng)絡(luò)模型。最后,將論文所提的DCNN模型應(yīng)用于航班延誤預(yù)測(cè)。在美國(guó)交通運(yùn)輸統(tǒng)計(jì)局提供的真實(shí)航班運(yùn)行數(shù)據(jù)上,DCNN模型預(yù)測(cè)航班延誤等級(jí)的準(zhǔn)確率為92.08%,有關(guān)成果可以為機(jī)場(chǎng)和旅客提供服務(wù)和指導(dǎo),具有非常重要的現(xiàn)實(shí)意義。
[Abstract]:As a branch of deep learning, convolutional neural networks have achieved great success in many fields, in which the depth of neural networks is the key to its success. However, the deeper the neural network is, the more difficult it is to train it. Therefore, this paper will focus on how to design and train a deep convolution neural network model. The main work of this paper is as follows: firstly, combining the characteristics of convolution neural network, a single-channel convolution neural network (SCNN) model is proposed, and its implementation and training flow are introduced in detail. In order to reduce the risk of model overfitting, Dropout algorithm is introduced into SCNN to improve the generalization ability of SCNN. In order to speed up the model training, batch Normalization BN (batch Normalization BN) algorithm is introduced into the SCNN model to normalize the activation values of all convolution layers in the network. Then, a fast and efficient Dual-Channel Convolution Neural Network (Dual-Channel Convolution Neural Network) model is proposed to solve the problem that the deep convolution neural network is difficult to train due to the disappearance of the gradient. The model consists of two channels: through channel and convolution channel. The through channel is responsible for ensuring the smooth flow of the deep network, and the convolution channel is responsible for the study of the deep network. Considering that the deep network is prone to unstable performance during training, the convolution attenuation factor is introduced into the convolution channel to constrain the response data. In order to ensure the consistency of each channel data dimension, a double cell layer is designed to de-sample the same feature map. On the CIFAR-10 CIFAR-100 and MNIST standard image recognition data sets, DCNN has achieved 94.53% and 99.74% classification accuracy, respectively. Both the training depth, stability and classification accuracy of the neural network are obviously superior to most of the existing deep convolution neural network models. Finally, the DCNN model proposed in this paper is applied to flight delay prediction. Based on the actual flight operation data provided by the United States Transportation Statistics Bureau, the accuracy of DCNN model in predicting flight delay grade is 92.08. The results can provide service and guidance for airports and passengers, which is of great practical significance.
【學(xué)位授予單位】：中國(guó)民航大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP311.13;TP18

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 劉凱;張立民;范曉磊;;改進(jìn)卷積玻爾茲曼機(jī)的圖像特征深度提取[J];哈爾濱工業(yè)大學(xué)學(xué)報(bào);2016年05期

2 李寰宇;畢篤彥;查宇飛;楊源;;一種易于初始化的類(lèi)卷積神經(jīng)網(wǎng)絡(luò)視覺(jué)跟蹤算法[J];電子與信息學(xué)報(bào);2016年01期

3 謝智歌;王岳青;竇勇;熊岳山;;基于卷積-自動(dòng)編碼機(jī)的三維形狀特征學(xué)習(xí)[J];計(jì)算機(jī)輔助設(shè)計(jì)與圖形學(xué)學(xué)報(bào);2015年11期

4 張晴晴;劉勇;潘接林;顏永紅;;基于卷積神經(jīng)網(wǎng)絡(luò)的連續(xù)語(yǔ)音識(shí)別[J];工程科學(xué)學(xué)報(bào);2015年09期

5 李寰宇;畢篤彥;楊源;查宇飛;覃兵;張立朝;;基于深度特征表達(dá)與學(xué)習(xí)的視覺(jué)跟蹤算法研究[J];電子與信息學(xué)報(bào);2015年09期

6 羅峗騫;陳志杰;湯錦輝;朱永文;;采用支持向量機(jī)回歸的航班延誤預(yù)測(cè)研究[J];交通運(yùn)輸系統(tǒng)工程與信息;2015年01期

7 許慶勇;江順亮;黃偉;李菁;徐少平;葉發(fā)茂;;基于多特征融合的深度置信網(wǎng)絡(luò)圖像分類(lèi)算法[J];計(jì)算機(jī)工程;2015年11期

8 潘煒深;金連文;馮子勇;;基于多尺度梯度及深度神經(jīng)網(wǎng)絡(luò)的漢字識(shí)別[J];北京航空航天大學(xué)學(xué)報(bào);2015年04期

9 程學(xué)旗;靳小龍;王元卓;郭嘉豐;張鐵贏;李國(guó)杰;;大數(shù)據(jù)系統(tǒng)和分析技術(shù)綜述[J];軟件學(xué)報(bào);2014年09期

10 余凱;賈磊;陳雨強(qiáng);徐偉;;深度學(xué)習(xí)的昨天、今天和明天[J];計(jì)算機(jī)研究與發(fā)展;2013年09期

相關(guān)博士學(xué)位論文 前1條

1 劉玉潔;基于貝葉斯網(wǎng)絡(luò)的航班延誤與波及預(yù)測(cè)[D];天津大學(xué);2009年

相關(guān)碩士學(xué)位論文 前1條

1 張孟;航班延誤引發(fā)旅客群體性事件處置對(duì)策優(yōu)化研究[D];山東財(cái)經(jīng)大學(xué);2016年

，

本文編號(hào)：2118545