def check_numerical_gradient():
X = loader.load_train_imgs()[:64, :10]
Y = loader.load_train_labels()[:10]
levels = [64, 20, 20, 10]
theta = initialization(levels)
print theta.shape
lbda = 3e-3
grad = J(theta, X, Y, levels, lbda)[1]
numeric_grad = compute_numerical_gradient(J, theta, X, Y, levels, lbda)
for i, g in enumerate(grad):
print g, numeric_grad[i]
def train():
global input_size, num_labels, hidden_size_l1, hidden_size_l2, lbda
levels = [input_size, hidden_size_l1, hidden_size_l2, num_labels]
train_data = loader.load_train_imgs()
train_labels = loader.load_train_labels()
WB_l1, WB_l2, sr_mat = pre_train(train_data, train_labels)
init_theta = stack_params((WB_l1[0], WB_l2[0]), sr_mat)
theta = fine_tuning(init_theta, train_data, train_labels, levels, lbda)
return theta
def test(theta):
global input_size, num_labels, hidden_size_l1, hidden_size_l2
levels = [input_size, hidden_size_l1, hidden_size_l2, num_labels]
train_data = loader.load_train_imgs(u'../MNISTHelper/t10k-images.idx3-ubyte')
train_labels = loader.load_train_labels(u'../MNISTHelper/t10k-labels.idx1-ubyte')
pl = predict(train_data, theta, levels)
print train_labels[:20]
print pl[:20]
e = 0.
for i in range(len(pl)):
if train_labels[i] != pl[i]:
e += 1
print 'error rate:', e/len(pl)
def main():
global input_size, num_labels, hidden_size, ro, lbda, beta
levels = [input_size, hidden_size, input_size]
mnist_data = loader.load_train_imgs()
mnist_labels = loader.load_train_labels()
labeled_set = np.where(mnist_labels < num_labels)[0]
unlabeled_set = np.where(mnist_labels > num_labels-1)[0]
train_num = np.round(len(labeled_set)/2)
train_data = mnist_data[:, labeled_set[:train_num]]
train_labels = mnist_labels[labeled_set[:train_num]]
test_data = mnist_data[:, labeled_set[train_num:]]
test_labels = mnist_labels[labeled_set[train_num:]]
unlabeled_data = mnist_data[:, unlabeled_set]
print "train data:", train_data.shape[1]
print "test data:", test_data.shape[1]
print "unlabeled data:", unlabeled_data.shape[1]
init_theta = initialization(levels)
theta, cost, info = fmin_l_bfgs_b(J, init_theta, args=(unlabeled_data, unlabeled_data, levels, lbda, ro, beta), approx_grad=False, maxiter=400)
print "cost:", cost
print "info:", info
WB = vec2mat(theta, levels)
sio.savemat('W', {'W': WB[0][0]})
train_a2 = feed_forward(WB[0], train_data, hidden_size)
test_a2 = feed_forward(WB[0], test_data, hidden_size)
sr_init_theta = sr_train.initialize_theta(hidden_size, num_labels)
sr_theta = sr_train.train(sr_init_theta, train_a2, train_labels, num_labels)
sr_mat = sr_vec2mat(sr_theta, num_labels)
pY = sr_predict(sr_mat, test_a2)
print test_labels[:20]
print pY[:20]
miss = 0.
for i, l in enumerate(test_labels):
if l != pY[i]:
miss += 1.
print miss/len(test_labels)
theta, cost, info = fmin_l_bfgs_b(J, init_theta, args=((X, Y, lbda, num_classes),), approx_grad=False, maxiter=400)
except Exception as e:
print e
print theta
print cost
print info
return theta
if __name__ == '__main__':
inputSize = 28*28
numClasses = 10
lbda = 1e-4
init_theta = initialize_theta(inputSize, numClasses)
X = load_train_imgs(u'../MNISThelper/train-images.idx3-ubyte')
Y = load_train_labels(u'../MNISThelper/train-labels.idx1-ubyte')
print X.shape, Y.shape, init_theta.shape
# numerical gradient check
'''
cX = X[:20, :10]
cY = Y[:10]
check_numerical_gradient(init_theta[:200], cX, cY, numClasses, lbda=lbda)
'''
theta = train(init_theta, X, Y, numClasses, lbda)
mat = vec2mat(theta, numClasses)
tX = load_train_imgs('../MNISThelper/t10k-images.idx3-ubyte')
tY = load_train_labels('../MNISThelper/t10k-labels.idx1-ubyte')
pY = predict(mat, tX)
print tY[:20]
print pY[:20]
e = 0.
