def get_data(N, N_test):
#load MNIST data using libraries available
training_data, training_labels = load_mnist('training')
test_data, test_labels = load_mnist('testing')
training_data = flatArray(N, 784,
training_data) #training_data is N x 784 matrix
training_labels = training_labels[:N]
test_data = flatArray(N_test, 784, test_data)
test_labels = test_labels[:N_test]
# adding column of 1s for bias
#training_data = addOnesColAtStart(training_data)
#test_data = addOnesColAtStart(test_data)
# Last 10% of training data size will be considered as the validation set
N_validation = int(N / 10)
validation_data = training_data[N - N_validation:N]
validation_labels = training_labels[N - N_validation:N]
N = N - N_validation
#update training data to remove validation data
training_data = training_data[:N]
training_labels = training_labels[:N]
return training_data, training_labels, test_data, test_labels, validation_data, validation_labels
def get_data(N=60000, N_test=10000, validationReqd=True):
# Load MNIST data using libraries available
training_data, training_labels = load_mnist('training')
test_data, test_labels = load_mnist('testing')
# Training_data is N x 784 matrix
training_data = flattenAndNormalize(N, 784, training_data)
training_labels = training_labels[:N]
test_data = flattenAndNormalize(N_test, 784, test_data)
test_labels = test_labels[:N_test]
# Adding column of 1s for bias
training_data = addOnesColAtStart(training_data)
test_data = addOnesColAtStart(test_data)
if (validationReqd):
# Last 10% of training data size will be considered as the validation set
N_validation = int(N / 6.0)
validation_data = training_data[N - N_validation:N]
validation_labels = training_labels[N - N_validation:N]
N = N - N_validation
else:
validation_data = []
validation_labels = []
#update training data to remove validation data
training_data = training_data[:N]
training_labels = training_labels[:N]
return training_data, training_labels, test_data, test_labels, validation_data, validation_labels
def get_data(split_percentage):
X_images, X_labels = load_mnist("training", np.arange(10), ".")
y_train = np.array(X_labels)
X_train = np.reshape(X_images, (60000, 784))
X_images, X_labels = load_mnist("testing", np.arange(10), ".")
y_test = np.array(X_labels)
X_test = np.reshape(X_images, (10000, 784))
print X_train.shape, X_test.shape, y_train.shape, y_test.shape
X_train = X_train / 255.
X_test = X_test / 255.
X = np.vstack((X_train, X_test))
y = np.concatenate((y_train, y_test))
#test_image=np.reshape(X_images[30],(28,28))
#plt.imshow(test_image,cmap='gray')
#plt.show()
#mnist = fetch_mldata("MNIST original")
#num_train_rows=split_percentage*70000/100
#num_test_rows=70000-num_train_rows
#X_train, X_test = X[:num_train_rows], X[num_train_rows:]
#y_train, y_test = y[:num_train_rows], y[num_train_rows:]
#return X_train,X_test,y_train,y_test
return X, y