卷积神经网络的编码文档指导及详细推导==.pptxVIP

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@1. Background2. Method neural network back-propagation3. Experiment training and testing samples neural network structure resultsNN digit recognitionbackgroundOCR (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 recognition , postcode recognition, the statistics of report forms and financial report forms. So the research on the Digit recognition is important.methodNeural network is a machine learning method.neural networkmethodBack PropagationBack propagation(BP) is a neural network learning algorithm.EnvironmentHardware: CPU: Intel core i7 3630QM, 2.4GHZ Memory: 8G ByteSoftware: OS: windows 7 (64) IDE: vs2010 (C++)ExperimentExperiment-Training and testing SamplesTraining and testing samples are fromMNIST ，/exdb/mnist//exdb/mnist/Every pic is a 28*28 dot matrixThere are 10 pics in a sample,and there are 891 samples ExperimentOur neural network has 3 layers：LayerInput layerHidden layerOutput layerDimension14*14eaningAverage grey value of four neighbor dotsEmpirical value , about 1~1.5 times of the inputTarget numbers , I.e. 0,1,2,3….9Neural network designExperimentBP neural networkneural network designExperimentresultsTraining samples：Testing samples：Training itemsSample picsRight predictionRight ratioTraining timevalue7000688898.4%603 476msTesting itemstest picsRight predictionRight ratioTraining timevalue1900179194.2632%908ms(0.48ms/pic)Experiment卷积神经网络基于人工神经网络在人工神经网络前，用滤波器进行特征抽取使用卷积核作为特征抽取器自动训练特征抽取器（即卷积核，即阈值参数）卷积卷积其实是一个图像处理核卷积用于增强图像的某种特征卷积的例子子采样降低图像分辨率减少训练维数增强网络对大小变化的适用性一般卷积神经网络的结构我的卷积神经网络结构实验效果训练时间次数及准确率问题1 准确率太低2 准确率抖动厉害3 单线程，训练速度太慢Thank you for listening!