【摘要】：隨著大數(shù)據(jù)時代的到來,計算機(jī)硬件性能的不斷提升,深度學(xué)習(xí)作為新興的機(jī)器學(xué)習(xí)方法被用于有效地分析和處理這些數(shù)據(jù)。深度學(xué)習(xí)的核心思想是采用一系列的非線性變換,從原始數(shù)據(jù)中提取由低層到高層、由一般到特定語義的特征。而卷積神經(jīng)網(wǎng)絡(luò)尤其擅長在高維復(fù)雜數(shù)據(jù)結(jié)構(gòu)中提取有效特征。正是這種豐富的特征表達(dá)能力使得卷積神經(jīng)網(wǎng)絡(luò)在圖像識別與分類、目標(biāo)檢測與定位、人機(jī)博弈、無人駕駛等領(lǐng)域應(yīng)用廣泛。TensorFlow是谷歌公司開源的深度學(xué)習(xí)平臺,也目前最受歡迎的機(jī)器學(xué)習(xí)框架。本文基于TensorFlow研究卷積神經(jīng)網(wǎng)絡(luò),并在此平臺基礎(chǔ)之上實現(xiàn)卷積神經(jīng)網(wǎng)絡(luò)模型,解決實際問題。具體工作如下:首先,對深度學(xué)習(xí)的基本方法進(jìn)行了介紹,重點研究了卷積神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)中的卷積層和池化層,并且搭建了TensorFlow實驗平臺,深刻理解TensorFlow的工作原理及框架結(jié)構(gòu)。其次,具體分析了 LeNet-5模型結(jié)構(gòu),使用兩個卷積層加一個全連接層構(gòu)建一個簡單的卷積神經(jīng)網(wǎng)絡(luò)解決手寫體數(shù)字識別問題,改進(jìn)后的LeNet-5模型在MNIST數(shù)據(jù)集上取得99.3%的準(zhǔn)確率。最后,對Alex描述的cuda-convnet模型使用了一些新的技巧進(jìn)行改進(jìn),主要是對weights進(jìn)行了 L2的正則化、對圖片進(jìn)行了翻轉(zhuǎn)隨機(jī)剪裁等數(shù)據(jù)增強(qiáng)以制造更多的樣本、在每個卷積-最大池化層后面使用了 LRN層以增強(qiáng)模型的泛化能力。改進(jìn)后的卷積神經(jīng)網(wǎng)絡(luò)在更復(fù)雜更豐富的CIFAR-10數(shù)據(jù)集上取得約88%的準(zhǔn)確率。
[Abstract]:With the arrival of big data era and the continuous improvement of computer hardware performance, depth learning as a new machine learning method is used to analyze and process these data effectively. The core idea of depth learning is to use a series of nonlinear transformations to extract features from lower level to higher level and from general to specific semantics from the original data. Convolutional neural networks are especially good at extracting effective features from high-dimensional complex data structures. It is this rich feature expression ability that makes convolutional neural network widely used in image recognition and classification, target detection and location, man-machine game, driverless and other fields. Tensor flow is Google's open source in-depth learning platform. Also currently the most popular machine learning framework. In this paper, the convolution neural network is studied based on TensorFlow, and the model of convolutional neural network is implemented on this platform to solve the practical problems. The main work is as follows: firstly, the basic method of deep learning is introduced, and the convolution layer and pool layer in the network structure of convolutional neural network are studied, and the TensorFlow experimental platform is built to deeply understand the working principle and frame structure of TensorFlow. Secondly, the structure of LeNet-5 model is analyzed in detail. A simple convolution neural network is constructed by using two convolution layers and a full join layer to solve the problem of handwritten digit recognition. The improved LeNet-5 model achieves 99.3% accuracy on MNIST data set. Finally, the cuda-convnet model described by Alex is improved with some new techniques, mainly the regularization of L2 for weights and the enhancement of image data such as flipping random clipping to create more samples. The LRN layer is used after each convolution-maximum pool layer to enhance the generalization of the model. The improved convolution neural network achieves an accuracy of about 88% on the more complex and abundant CIFAR-10 datasets.
【學(xué)位授予單位】：華中師范大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41;TP18

【參考文獻(xiàn)】

相關(guān)博士學(xué)位論文 前7條

1 戶保田;基于深度神經(jīng)網(wǎng)絡(luò)的文本表示及其應(yīng)用[D];哈爾濱工業(yè)大學(xué);2016年

2 馮子勇;基于深度學(xué)習(xí)的圖像特征學(xué)習(xí)和分類方法的研究及應(yīng)用[D];華南理工大學(xué);2016年

3 吳翔;基于機(jī)器視覺的害蟲識別方法研究[D];浙江大學(xué);2016年

4 蘇向東;基于深度學(xué)習(xí)和知識策略的蒙古文古籍識別研究[D];內(nèi)蒙古大學(xué);2016年

5 謝凌曦;基于局部特征的圖像表示模型理論與實踐[D];清華大學(xué);2015年

6 張清辰;面向大數(shù)據(jù)特征學(xué)習(xí)的深度計算模型研究[D];大連理工大學(xué);2015年

7 馮方向;基于深度學(xué)習(xí)的跨模態(tài)檢索研究[D];北京郵電大學(xué);2015年

相關(guān)碩士學(xué)位論文 前5條

1 吳正文;卷積神經(jīng)網(wǎng)絡(luò)在圖像分類中的應(yīng)用研究[D];電子科技大學(xué);2015年

2 張勇;深度卷積神經(jīng)網(wǎng)絡(luò)在車牌和人臉檢測領(lǐng)域的應(yīng)用研究[D];鄭州大學(xué);2015年

3 何西麟;基于深度學(xué)習(xí)的手寫體字符識別研究與實現(xiàn)[D];中山大學(xué);2015年

4 陳先昌;基于卷積神經(jīng)網(wǎng)絡(luò)的深度學(xué)習(xí)算法與應(yīng)用研究[D];浙江工商大學(xué);2014年

5 許可;卷積神經(jīng)網(wǎng)絡(luò)在圖像識別上的應(yīng)用的研究[D];浙江大學(xué);2012年

，

本文編號：2227939