神經(jīng)網(wǎng)絡(luò)及卷積神經(jīng)網(wǎng)絡(luò)用于手寫(xiě)數(shù)字識(shí)別

1、A Hand-writing digit recognition Application base on Neural network and covolutional neural networkColleague of Computer ScienceInner Mongolia UniversityHohhot , Inner Mongolia , P.R ChinaEmail: huangzhiqiang共二十一頁(yè)NN digit recognition1. Background2. Method neural network back-propagation3. Experiment

2、 training and testing samples neural network structure results共二十一頁(yè)OCR (Optical Character Recognition) has become one of the important methods in gathering information and information transformation. Digit recognition has a promising feature in many fields in society, such as the car license plate r

3、ecognition , postcode recognition, the statistics of report forms and financial report forms. So the research on the Digit recognition is important.background共二十一頁(yè)neural networkNeural network is a machine learning method.method共二十一頁(yè)Back PropagationBack propagation(BP) is a neural network learning al

4、gorithm.method共二十一頁(yè)EnvironmentHardware: CPU: Intel core i7 3630QM, 2.4GHZ Memory: 8G ByteSoftware: OS: windows 7 (64) IDE: vs2010 (C+)Experiment共二十一頁(yè)Experiment-Training and testing SamplesTraining and testing samples are fromMNIST ，/exdb/mnist/Every pic is a 28*28 dot matrixThere are 10 pics in a sa

5、mple,and there are 891 samples Experiment共二十一頁(yè)Neural network designOur neural network has 3 layers：LayerInput layerHidden layerOutput layerDimension14*14=19619610meaningAverage grey value of four neighbor dotsEmpirical value , about 11.5 times of the inputTarget numbers , I.e. 0,1,2,3.9Experiment共二十

6、一頁(yè)neural network designBP neural networkExperiment共二十一頁(yè)resultsTraining samples：Testing samples：Training itemsSample picsRight predictionRight ratioTraining timevalue7000688898.4%603 476msTesting itemstest picsRight predictionRight ratioTraining timevalue1900179194.2632%908ms(0.48ms/pic)Experiment共二十

7、一頁(yè)卷積神經(jīng)網(wǎng)絡(luò)基于人工神經(jīng)網(wǎng)絡(luò)在人工神經(jīng)網(wǎng)絡(luò)前，用濾波器進(jìn)行特征抽取(chu q)使用卷積核作為特征抽取器自動(dòng)訓(xùn)練特征抽取器（即卷積核，即閾值參數(shù)）共二十一頁(yè)卷積卷積其實(shí)(qsh)是一個(gè)圖像處理核卷積用于增強(qiáng)圖像的某種特征共二十一頁(yè)卷積的例子(l zi)共二十一頁(yè)子采樣(ci yn)降低圖像分辨率減少(jinsho)訓(xùn)練維數(shù)增強(qiáng)網(wǎng)絡(luò)對(duì)大小變化的適用性共二十一頁(yè)一般(ybn)卷積神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)共二十一頁(yè)我的卷積神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)(jigu)共二十一頁(yè)實(shí)驗(yàn)(shyn)效果共二十一頁(yè)訓(xùn)練(xnlin)時(shí)間次數(shù)及準(zhǔn)確率共二十一頁(yè)問(wèn)題(wnt)1 準(zhǔn)確率太低2 準(zhǔn)確率抖動(dòng)厲害3 單線(xiàn)程，訓(xùn)練(xnlin)速度太慢共二十一頁(yè)Thank you for listening!共二十一頁(yè)內(nèi)容摘要A Hand-writing digit recognition Application base on Neural network and covolutional neural network。A Hand-writing digit recognition Application base on Neural network and covoluti