本文選題：卷積神經(jīng)網(wǎng)絡(luò) + 激活函數(shù)　； 參考：《北京交通大學(xué)》2017年碩士論文

【摘要】：卷積神經(jīng)網(wǎng)絡(luò)屬于深度學(xué)習(xí)領(lǐng)域研究的范圍,是一種高效的識(shí)別方法,卷積神經(jīng)網(wǎng)絡(luò)具有三個(gè)特點(diǎn)分別為參數(shù)共享,局部感知和子采樣操作,這三個(gè)特點(diǎn)使得訓(xùn)練參數(shù)減少,訓(xùn)練速度加快,在訓(xùn)練過程中具有良好表現(xiàn),目前卷積神經(jīng)網(wǎng)絡(luò)已經(jīng)廣泛的并且良好的應(yīng)用在生活各個(gè)方面,特別是在圖像分類任務(wù),語(yǔ)音識(shí)別,文本識(shí)別,路標(biāo)識(shí)別等方面。但其發(fā)展過程中還存在一些問題。本文將對(duì)卷積神經(jīng)網(wǎng)絡(luò)在圖像分類領(lǐng)域進(jìn)行研究,目的是希望提高圖像分類的精準(zhǔn)率,降低錯(cuò)誤率。激活函數(shù)通過非線性函數(shù)把激活的神經(jīng)元的特征保留并映射出來,因此對(duì)于網(wǎng)絡(luò)性能有很大的影響,但是目前激活函數(shù)的選擇是一個(gè)問題,不同的激活函數(shù)具有不同的優(yōu)缺點(diǎn),需要耗費(fèi)大量的時(shí)間與精力來確定最優(yōu)的激活函數(shù)。本文主要針對(duì)激活函數(shù)選擇困難的問題,提出基于Relu-Softplus激活函數(shù)的卷積神經(jīng)網(wǎng)絡(luò),并在手寫數(shù)字字體MNIST數(shù)據(jù)集上進(jìn)行實(shí)驗(yàn),加以驗(yàn)證其性能,并且同其他不同的激活函數(shù)進(jìn)行比對(duì),分析其圖像分類的錯(cuò)誤率,以及收斂速度的快慢,最終達(dá)到優(yōu)化卷積神經(jīng)網(wǎng)絡(luò)的性能和解決確定最優(yōu)激活函數(shù)困難等問題的目的。卷積神經(jīng)網(wǎng)絡(luò)中的學(xué)習(xí)方式常見的有兩種,有監(jiān)督學(xué)習(xí)方法和無監(jiān)督學(xué)習(xí)方法,有監(jiān)督學(xué)習(xí)即從已標(biāo)記的訓(xùn)練樣本中學(xué)習(xí)到映射函數(shù),但是需要大量的訓(xùn)練樣本,并且易出現(xiàn)過擬合等問題。而無監(jiān)督學(xué)習(xí)不要求訓(xùn)練樣本帶有標(biāo)簽,希望學(xué)習(xí)到更過抽象隱藏的特征結(jié)構(gòu),但具有訓(xùn)練時(shí)間長(zhǎng),訓(xùn)練過程繁瑣等缺點(diǎn)。本文主要針對(duì)此問題,提出基于K-means算法的卷積神經(jīng)網(wǎng)絡(luò),并在CIFAR-10數(shù)據(jù)集上進(jìn)行實(shí)驗(yàn),加以驗(yàn)證其性能,并分析比較不同的網(wǎng)絡(luò)框架對(duì)圖像分類精準(zhǔn)率的影響。最后本論文將卷積神經(jīng)網(wǎng)絡(luò)應(yīng)用在路標(biāo)識(shí)別系統(tǒng)上,并且設(shè)計(jì)了一個(gè)路標(biāo)識(shí)別系統(tǒng),從系統(tǒng)的需求分析,概要設(shè)計(jì),詳細(xì)設(shè)計(jì)以實(shí)現(xiàn)等方面進(jìn)行了闡述。并將本文提出的基于K-means算法的卷積神經(jīng)網(wǎng)絡(luò)應(yīng)用在路標(biāo)識(shí)別系統(tǒng)中,最后在德國(guó)交通標(biāo)志識(shí)別GTRSB數(shù)據(jù)集上進(jìn)行訓(xùn)練測(cè)試,并同其他知名的算法進(jìn)行比較,加以驗(yàn)證了基于K-means算法的卷積神經(jīng)網(wǎng)絡(luò)在路標(biāo)識(shí)別系統(tǒng)的應(yīng)用中對(duì)于路標(biāo)分類的準(zhǔn)確性,可靠性以及時(shí)效性方面確實(shí)有一定的提升。
[Abstract]:Convolutional neural network is an efficient recognition method, which belongs to the field of deep learning. It has three characteristics: parameter sharing, local sensing and sub-sampling operation, which make the training parameters reduced. At present, convolution neural network has been widely used in all aspects of life, especially in image classification task, speech recognition, text recognition, road sign recognition and so on. However, there are still some problems in its development. In this paper convolution neural networks are studied in the field of image classification in order to improve the accuracy of image classification and reduce the error rate. The activation function preserves and maps the characteristics of the activated neuron through the nonlinear function, so it has a great influence on the network performance. But at present, the choice of the activation function is a problem, and different activation functions have different advantages and disadvantages. It takes a lot of time and effort to determine the optimal activation function. Aiming at the difficulty of selecting activation function, a convolutional neural network based on Relu-Softplus activation function is proposed in this paper. Experiments are carried out on the MNIST dataset of handwritten digital font to verify its performance. Compared with other activation functions, the error rate of image classification and the speed of convergence are analyzed. Finally, the performance of convolution neural network is optimized and the problem of determining the optimal activation function is solved. There are two common learning methods in convolutional neural networks: supervised learning and unsupervised learning. Supervised learning is learning mapping functions from marked training samples, but a large number of training samples are required. And easy to have problems such as fitting. But the unsupervised learning does not require the training samples to be labeled, hoping to learn more abstract and hidden feature structures, but it has the disadvantages of long training time and tedious training process. In order to solve this problem, a convolutional neural network based on K-means algorithm is proposed in this paper. Experiments are carried out on the CIFAR-10 dataset to verify its performance, and the effects of different network frameworks on the accuracy rate of image classification are analyzed and compared. Finally, this paper applies the convolution neural network to the signpost recognition system, and designs a signpost recognition system, which is described from the aspects of system requirement analysis, summary design, detailed design and so on. The convolutional neural network based on K-means algorithm is applied to the road sign recognition system. Finally, the training test is carried out on the GTRSB data set of traffic sign recognition in Germany, and compared with other well-known algorithms. It is verified that convolution neural network based on K-means algorithm can improve the accuracy, reliability and timeliness of road sign classification in the application of road sign recognition system.
【學(xué)位授予單位】：北京交通大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41;TP183

【參考文獻(xiàn)】

相關(guān)期刊論文 前6條

1 黃毅;段修生;孫世宇;郎巍;;基于改進(jìn)sigmoid激活函數(shù)的深度神經(jīng)網(wǎng)絡(luò)訓(xùn)練算法研究[J];計(jì)算機(jī)測(cè)量與控制;2017年02期

2 趙玲玲;楊輝華;劉振丙;潘細(xì)朋;;基于深度卷積神經(jīng)網(wǎng)絡(luò)的乳腺細(xì)胞圖像分類研究[J];中小企業(yè)管理與科技(下旬刊);2016年06期

3 孫艷豐;楊新東;胡永利;王萍;;基于Softplus激活函數(shù)和改進(jìn)Fisher判別的ELM算法[J];北京工業(yè)大學(xué)學(xué)報(bào);2015年09期

4 呂剛;郝平;盛建榮;;一種改進(jìn)的深度神經(jīng)網(wǎng)絡(luò)在小圖像分類中的應(yīng)用研究[J];計(jì)算機(jī)應(yīng)用與軟件;2014年04期

5 趙雷;張延榮;;基于概率神經(jīng)網(wǎng)絡(luò)和K-means算法的納稅評(píng)估[J];河北工程大學(xué)學(xué)報(bào)(社會(huì)科學(xué)版);2011年01期

6 吳佑壽,趙明生;激活函數(shù)可調(diào)的神經(jīng)元模型及其有監(jiān)督學(xué)習(xí)與應(yīng)用[J];中國(guó)科學(xué)E輯:技術(shù)科學(xué);2001年03期

相關(guān)碩士學(xué)位論文 前8條

1 姜含露;基于卷積神經(jīng)網(wǎng)的高光譜數(shù)據(jù)特征提取及分類技術(shù)研究[D];哈爾濱工業(yè)大學(xué);2016年

2 產(chǎn)文濤;基于卷積神經(jīng)網(wǎng)絡(luò)的人臉表情和性別識(shí)別[D];安徽大學(xué);2016年

3 何云超;聚類算法和卷積神經(jīng)網(wǎng)絡(luò)在文本情感分析中的應(yīng)用研究[D];云南大學(xué);2016年

4 張興革;基于卷積神經(jīng)網(wǎng)絡(luò)模型下的語(yǔ)音處理方法研究[D];東北林業(yè)大學(xué);2016年

5 楊楠;基于Caffe深度學(xué)習(xí)框架的卷積神經(jīng)網(wǎng)絡(luò)研究[D];河北師范大學(xué);2016年

6 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

7 岳永鵬;深度無監(jiān)督學(xué)習(xí)算法研究[D];西南石油大學(xué);2015年

8 張凱歌;基于K-means和神經(jīng)網(wǎng)絡(luò)算法的圖像文字提取與識(shí)別[D];云南大學(xué);2013年

，

本文編號(hào)：1775377