training_set = train
testing_set = test
else:
training_set = np.c_[training_set, train]
testing_set = np.c_[testing_set, test]
labels = 10 * [index]
testing_lables += labels
training_labels += 8 * labels
testing_lables = np.array(testing_lables)
training_labels = np.array(training_labels)
training_data = samplingImg(training_set)
# step2 L1 feature learning using sparse autoencoder
W = nnet.sparseAutoencoder(inputSize, hiddenSizeL1, sparsityParam, lmd, beta, alpha, training_data)
W1 = np.reshape(W[: hiddenSizeL1 * inputSize,], (hiddenSizeL1, inputSize))
b1 = np.reshape(W[2 * hiddenSizeL1 * inputSize : 2 * hiddenSizeL1 * inputSize + hiddenSizeL1,], (hiddenSizeL1, 1))
# step3 convolution layer, compute feature map
# TODO extract imagesize
step = 1
imagesize = 50
convWeight = convolutionWeight(W1, patchsize, imagesize, step)
featureMap = convolutionFeatureMap(training_set, b1, convWeight)
# step4 pooling layer
poolingSize = 2
poolingCore = 1.0 / math.pow(poolingSize, 2) * np.ones((1, poolingSize * poolingSize))
featureSize = math.sqrt(featureMap[0].shape[0])
poolingWeight = convolutionWeight(poolingCore, poolingSize, featureSize, poolingSize)
poolingWeight = poolingWeight[0]
#step2 divide training set into feature learning set & supervised training set
labeled_set = np.nonzero(labels>=5)
unlabeled_set = np.nonzero(labels<=4)
labeled_dataset = data[:,(labels<=4).flatten()]
labeled_labelset = labels[(labels<=4).flatten()]
unlabeled_dataset = data[:,(labels>=5).flatten()]
unlabeled_dataset = unlabeled_dataset[:,:unlabeled_dataset.shape[1]/3]
print unlabeled_dataset.shape
#step3 feature learning using sparse autoencoder
#TODO move normalization to miscellaneous
unlabeled_dataset = nnet.normalization(unlabeled_dataset)
W = nnet.sparseAutoencoder(inputSize,hiddenSize,sparsityParam,lmd,beta,alpha,unlabeled_dataset)
W1 = np.reshape(W[:hiddenSize*inputSize,], (hiddenSize, inputSize))
b1 = np.reshape(W[2*hiddenSize*inputSize:2*hiddenSize*inputSize+hiddenSize,],(hiddenSize,1))
#step4 plot the learned feature
fig = plt.figure(2)
for index in range(hiddenSize/10):
weight = W1[index,:]
weight = np.reshape(weight,(28,28))
ax = fig.add_subplot(5,4,1+index)
ax.imshow(weight,mpl.cm.gray)
plt.show()
#step5 extract features from test & training data
training_set = train
testing_set = test
else:
training_set = np.c_[training_set, train]
testing_set = np.c_[testing_set, test]
labels = 10*[index]
testing_lables += labels
training_labels += 8*labels
testing_lables = np.array(testing_lables)
training_labels = np.array(training_labels)
training_data = samplingImg(training_set)
#step2 L1 feature learning using sparse autoencoder
W = nnet.sparseAutoencoder(inputSize,hiddenSizeL1,sparsityParam,lmd,beta,alpha,training_data)
W1 = np.reshape(W[:hiddenSizeL1*inputSize,], (hiddenSizeL1, inputSize))
b1 = np.reshape(W[2*hiddenSizeL1*inputSize:2*hiddenSizeL1*inputSize+hiddenSizeL1,],(hiddenSizeL1,1))
#step3 convolution layer, compute feature map
#TODO extract imagesize
step =1
imagesize = 50
convWeight = convolutionWeight(W1, patchsize, imagesize, step)
featureMap = convolutionFeatureMap(training_set, b1, convWeight)
#step4 pooling layer
poolingSize = 2
poolingCore = 1.0/math.pow(poolingSize, 2) * np.ones((1, poolingSize*poolingSize))
featureSize = math.sqrt(featureMap[0].shape[0])
poolingWeight = convolutionWeight(poolingCore, poolingSize, featureSize, poolingSize)
poolingWeight = poolingWeight[0]
