#step2 获取无标签数据集, 有标签训练数据集, 有标签测试数据集
unlabeledData, trainData, trainLabels, testData, testLabels = GetDivideSets(
)
#step3 用无标签数据集训练自编码的特征
theta = sparseAutoencoder.initiallize(hiddenSize, inputSize)
[X,cost,d]=sop.fmin_l_bfgs_b(lambda (x) :sparseAutoencoder.sparseAutoencoderCost(x, inputSize, \
hiddenSize, la, sparsityParam, beta, unlabeledData),x0=theta,maxiter=400,disp=1)
W1 = X[0:hiddenSize * inputSize].reshape(hiddenSize, inputSize)
W1 = W1.T
opttheta = X
#step4 获得有标签训练集的激活值并用来训练softmax分类器的权值
trainImages = feedforward(opttheta, hiddenSize, inputSize, trainData)
thetaSoftmax = softmax.initiallize(numClasses, hiddenSize)
la = 1e-4
[X, cost, d] = sop.fmin_l_bfgs_b(lambda (x): softmax.SoftmaxCost(
trainImages, trainLabels, x, la, hiddenSize, numClasses),
x0=thetaSoftmax,
maxiter=100,
disp=1)
#step5 获得有标签测试集的激活值并且给出softmax分类准确率, 5分类的准确率应该在98%附近
testImages = feedforward(opttheta, hiddenSize, inputSize, testData)
optthetaSoftmax = X
accuracy = softmax.predict(optthetaSoftmax, testImages, testLabels,
hiddenSize, numClasses)
print accuracy
#画出权值图像
hiddenSize = 200
sparsityParam = 0.1
la = 3e-3
beta = 3
#step2 获取无标签数据集, 有标签训练数据集, 有标签测试数据集
unlabeledData, trainData, trainLabels, testData, testLabels = GetDivideSets()
#step3 用无标签数据集训练自编码的特征
theta = sparseAutoencoder.initiallize(hiddenSize, inputSize)
[X,cost,d]=sop.fmin_l_bfgs_b(lambda (x) :sparseAutoencoder.sparseAutoencoderCost(x, inputSize, \
hiddenSize, la, sparsityParam, beta, unlabeledData),x0=theta,maxiter=400,disp=1)
W1 = X[0:hiddenSize*inputSize].reshape(hiddenSize, inputSize)
W1 = W1.T
opttheta = X
#step4 获得有标签训练集的激活值并用来训练softmax分类器的权值
trainImages = feedforward(opttheta, hiddenSize, inputSize, trainData)
thetaSoftmax = softmax.initiallize(numClasses, hiddenSize)
la = 1e-4
[X,cost,d] = sop.fmin_l_bfgs_b(lambda (x) :softmax.SoftmaxCost(trainImages, trainLabels,x,la,hiddenSize,numClasses),x0=thetaSoftmax,maxiter=100,disp=1)
#step5 获得有标签测试集的激活值并且给出softmax分类准确率, 5分类的准确率应该在98%附近
testImages = feedforward(opttheta, hiddenSize, inputSize, testData)
optthetaSoftmax = X
accuracy = softmax.predict(optthetaSoftmax,testImages,testLabels,hiddenSize,numClasses)
print accuracy
#画出权值图像
sparseAutoencoder.showimg(W1, hiddenSize, 28)
pooledFeaturesThis = cnnPool(poolDim, convolvedFeaturesThis)
pooledFeaturesTrain[featureStart:featureEnd, :, :, :] = pooledFeaturesThis
convolvedFeaturesThis = cnnConvolve(patchDim, stepSize, testImages, Wt, bt, white, meanPatch)
pooledFeaturesThis = cnnPool(poolDim, convolvedFeaturesThis)
pooledFeaturesTest[featureStart:featureEnd, :, :, :] = pooledFeaturesThis
cnnPooledFeatures = {}
cnnPooledFeatures['pooledFeaturesTrain'] = pooledFeaturesTrain
cnnPooledFeatures['pooledFeaturesTest'] = pooledFeaturesTest
sio.savemat('cnnPooledFeatures.mat', cnnPooledFeatures)
#step7 用卷积特征来进行softmax预测,这个就很快了,十几秒, 正确率为80%左右
#Mat1 = sio.loadmat('cnnPooledFeatures.mat')
#pooledFeaturesTrain = Mat1['pooledFeaturesTrain']
#pooledFeaturesTest = Mat1['pooledFeaturesTest']
softmaxLambda = 1e-4
numClasses = 4
inputSize = int(pooledFeaturesTrain.size/numTrainImages)
print inputSize
softmaxX = np.transpose(pooledFeaturesTrain, (0, 2, 3, 1))
softmaxX = softmaxX.reshape(inputSize, numTrainImages)
softmaxY = trainLabels
theta = softmax.initiallize(numClasses, inputSize)
[X,cost,d] = sop.fmin_l_bfgs_b(lambda (x) :softmax.SoftmaxCost(softmaxX, softmaxY, x, softmaxLambda, inputSize, numClasses),x0=theta,maxiter=200,disp=1)
softmaxX = np.transpose(pooledFeaturesTest, (0, 2, 3, 1))
softmaxX = softmaxX.reshape(inputSize, numTestImages)
softmaxY = testLabels
accuracy = softmax.predict(X,softmaxX,softmaxY,inputSize,numClasses)
print accuracy
#step3 训练第一层神经网络的权值
[sae1OptTheta, cost, d]=sop.fmin_l_bfgs_b(lambda (x) :sparseAutoencoder.sparseAutoencoderCost(x, inputSize, \
hiddenSizeL1, la, sparsityParam, beta, trainData),x0=sae1Theta,maxiter=400,disp=1)
sae1Features = feedforward(sae1OptTheta, hiddenSizeL1, inputSize,
trainData)
#step4 训练第二层神经网络的权值
sae2Theta = sparseAutoencoder.initiallize(hiddenSizeL2, hiddenSizeL1)
[sae2OptTheta, cost, d]=sop.fmin_l_bfgs_b(lambda (x) : sparseAutoencoder.sparseAutoencoderCost(x, hiddenSizeL1, \
hiddenSizeL2, la, sparsityParam, beta, sae1Features),x0=sae2Theta, maxiter=400,disp=1)
sae2Features = feedforward(sae2OptTheta, hiddenSizeL2, hiddenSizeL1,
sae1Features)
#step5 训练softmax权值
saeSoftmaxTheta = softmax.initiallize(numClasses, hiddenSizeL2)
laSoftmax = 1e-4
[saeSoftmaxOptTheta, cost, d] = sop.fmin_l_bfgs_b(lambda (x) :softmax.SoftmaxCost(sae2Features, \
trainLabels, x, laSoftmax, hiddenSizeL2, numClasses), x0=saeSoftmaxTheta, maxiter=100, disp=1)
#step6 微调栈式神经网络的所有权值
stackedAETheta, netconfig = initaillizeAETheta(sae1OptTheta, sae2OptTheta,
saeSoftmaxOptTheta,
hiddenSizeL1, hiddenSizeL2,
inputSize)
[stackedAEOptTheta, cost, d] = sop.fmin_l_bfgs_b(lambda (x):stackedAECost(x, inputSize, hiddenSizeL2, \
numClasses, netconfig, la, trainData, trainLabels), x0 = stackedAETheta, maxiter = 400, disp=1)
#step7 给出栈式神经网络的10分类的准确率, 微调前的应该在90%左右, 微调后在97%左右
accuracyBeforeFine = predict(stackedAETheta, inputSize, hiddenSizeL2,
numClasses, netconfig, testData, testLabels)
trainLabels = trainLabels[0:length]
sae1Theta = sparseAutoencoder.initiallize(hiddenSizeL1, inputSize)
#step3 训练第一层神经网络的权值
[sae1OptTheta, cost, d]=sop.fmin_l_bfgs_b(lambda (x) :sparseAutoencoder.sparseAutoencoderCost(x, inputSize, \
hiddenSizeL1, la, sparsityParam, beta, trainData),x0=sae1Theta,maxiter=400,disp=1)
sae1Features = feedforward(sae1OptTheta, hiddenSizeL1, inputSize, trainData)
#step4 训练第二层神经网络的权值
sae2Theta = sparseAutoencoder.initiallize(hiddenSizeL2, hiddenSizeL1)
[sae2OptTheta, cost, d]=sop.fmin_l_bfgs_b(lambda (x) : sparseAutoencoder.sparseAutoencoderCost(x, hiddenSizeL1, \
hiddenSizeL2, la, sparsityParam, beta, sae1Features),x0=sae2Theta, maxiter=400,disp=1)
sae2Features = feedforward(sae2OptTheta, hiddenSizeL2, hiddenSizeL1, sae1Features)
#step5 训练softmax权值
saeSoftmaxTheta = softmax.initiallize(numClasses, hiddenSizeL2)
laSoftmax = 1e-4
[saeSoftmaxOptTheta, cost, d] = sop.fmin_l_bfgs_b(lambda (x) :softmax.SoftmaxCost(sae2Features, \
trainLabels, x, laSoftmax, hiddenSizeL2, numClasses), x0=saeSoftmaxTheta, maxiter=100, disp=1)
#step6 微调栈式神经网络的所有权值
stackedAETheta, netconfig = initaillizeAETheta(sae1OptTheta, sae2OptTheta, saeSoftmaxOptTheta, hiddenSizeL1, hiddenSizeL2, inputSize)
[stackedAEOptTheta, cost, d] = sop.fmin_l_bfgs_b(lambda (x):stackedAECost(x, inputSize, hiddenSizeL2, \
numClasses, netconfig, la, trainData, trainLabels), x0 = stackedAETheta, maxiter = 400, disp=1)
#step7 给出栈式神经网络的10分类的准确率, 微调前的应该在90%左右, 微调后在97%左右
accuracyBeforeFine = predict(stackedAETheta, inputSize, hiddenSizeL2, numClasses, netconfig, testData, testLabels)
accuracyAfterFine = predict(stackedAEOptTheta, inputSize, hiddenSizeL2, numClasses, netconfig, testData, testLabels)
print 'accuracyBeforeFine: '+str(accuracyBeforeFine)
print 'accuracyAfterFine: '+str(accuracyAfterFine)
featureStart:featureEnd, :, :, :] = pooledFeaturesThis
cnnPooledFeatures = {}
cnnPooledFeatures['pooledFeaturesTrain'] = pooledFeaturesTrain
cnnPooledFeatures['pooledFeaturesTest'] = pooledFeaturesTest
sio.savemat('cnnPooledFeatures.mat', cnnPooledFeatures)
#step7 用卷积特征来进行softmax预测,这个就很快了,十几秒, 正确率为80%左右
#Mat1 = sio.loadmat('cnnPooledFeatures.mat')
#pooledFeaturesTrain = Mat1['pooledFeaturesTrain']
#pooledFeaturesTest = Mat1['pooledFeaturesTest']
softmaxLambda = 1e-4
numClasses = 4
inputSize = int(pooledFeaturesTrain.size / numTrainImages)
print inputSize
softmaxX = np.transpose(pooledFeaturesTrain, (0, 2, 3, 1))
softmaxX = softmaxX.reshape(inputSize, numTrainImages)
softmaxY = trainLabels
theta = softmax.initiallize(numClasses, inputSize)
[X, cost, d] = sop.fmin_l_bfgs_b(lambda (x): softmax.SoftmaxCost(
softmaxX, softmaxY, x, softmaxLambda, inputSize, numClasses),
x0=theta,
maxiter=200,
disp=1)
softmaxX = np.transpose(pooledFeaturesTest, (0, 2, 3, 1))
softmaxX = softmaxX.reshape(inputSize, numTestImages)
softmaxY = testLabels
accuracy = softmax.predict(X, softmaxX, softmaxY, inputSize, numClasses)
print accuracy