MNIST 被喻为深度学习中的Hello World示例，由Yann LeCun等大神组织收集的一个手写数字的数据集，有60000个训练集和10000个验证集，是个非常适合初学者入门的训练集。这个网站也提供了业界对这个数据集的各种算法的尝试结果，也能看出机器学习的算法的演进史，从早期的线性逻辑回归到K-means，再到两层神经网络，到多层神经网络，再到最近的卷积神经网络，随着的算法模型的改善，错误率也不断下降，所以目前这个数据集的错误率已经可以控制在0.2%左右，基本和人类识别的能力相当了。
这篇文章的例子我们会用一个更加有趣点的数据集 notMNIST，和MNIST不同的是它是一个各种形态的字母的数据集合，总共有a~j 10个字母组成，字母a相对应的一些图片如下：
在这个例子中，我们会使用TensorFlow和sklearn等库，对数据集进行一系列处理，最终使用逻辑回归模型来进行机器学习并且预测。
1. 准备环境
安装Python2.7和pip
Python2.7的官方网站：https://www.python.org/getit/
pip是Python Package Index，通过pip可以非常方便的查找安装其他软件，安装pip的方法如下：https://pip.pypa.io/en/stable/installing/
安装TensorFlow
$ pip install tensorflow
2. 下载数据
import需要的第三方库
# These are all the modules we'll be using later. Make sure you can import them
# before proceeding further.
from __future__ import print_function
import matplotlib.pyplot as plt
import numpy as np
import os
import sys
import tarfile
from IPython.display import display, Image
from scipy import ndimage
from sklearn.linear_model import LogisticRegression
from six.moves.urllib.request import urlretrieve
from six.moves import cPickle as pickle
# Config the matplotlib backend as plotting inline in IPython
%matplotlib inline
首先，我们会下载数据集到本地电脑。所有的图片都是28*28像素的图片，标示为"A"到"J"（10个分类）。整个数据集合包含大概50000个训练数据和19000个测试数据，所以这样规模的数据集合可以在大多数电脑上较快的完成训练。训练数据文件名是notMNIST_large.tar.gz，测试数据文件名是notMNIST_small.tar.gz。
url = 'http://commondatastorage.googleapis.com/books1000/'
last_percent_reported = None
data_root = '.' # Change me to store data elsewhere
def download_progress_hook(count, blockSize, totalSize):
"""A hook to report the progress of a download. This is mostly intended for users with
slow internet connections. Reports every 5% change in download progress.
"""
global last_percent_reported
percent = int(count * blockSize * 100 / totalSize)
if last_percent_reported != percent:
if percent % 5 == 0:
sys.stdout.write("%s%%" % percent)
sys.stdout.flush()
else:
sys.stdout.write(".")
sys.stdout.flush() last_percent_reported = percent def maybe_download(filename, expected_bytes, force=False):
"""Download a file if not present, and make sure it's the right size."""
dest_filename = os.path.join(data_root, filename)
if force or not os.path.exists(dest_filename):
print('Attempting to download:', filename)
filename, _ = urlretrieve(url + filename, dest_filename, reporthook=download_progress_hook)
print('\nDownload Complete!')
statinfo = os.stat(dest_filename)
if statinfo.st_size == expected_bytes:
print('Found and verified', dest_filename)
else:
raise Exception(
'Failed to verify ' + dest_filename + '. Can you get to it with a browser?')
return dest_filename
train_filename = maybe_download('notMNIST_large.tar.gz', 247336696)
test_filename = maybe_download('notMNIST_small.tar.gz', 8458043)
解压数据集合，会产生一系列标记从A到J的目录。
num_classes = 10
np.random.seed(133)
def maybe_extract(filename, force=False):
root = os.path.splitext(os.path.splitext(filename)[0])[0] # remove .tar.gz
if os.path.isdir(root) and not force:
# You may override by setting force=True.
print('%s already present - Skipping extraction of %s.' % (root, filename))
else:
print('Extracting data for %s. This may take a while. Please wait.' % root)
tar = tarfile.open(filename)
sys.stdout.flush()
tar.extractall(data_root)
tar.close()
data_folders = [
os.path.join(root, d) for d in sorted(os.listdir(root))
if os.path.isdir(os.path.join(root, d))]
if len(data_folders) != num_classes:
raise Exception(
'Expected %d folders, one per class. Found %d instead.' % (
num_classes, len(data_folders)))
print(data_folders)
return data_folders train_folders = maybe_extract(train_filename)
test_folders = maybe_extract(test_filename)
输出如下：
./notMNIST_large already present - Skipping extraction of ./notMNIST_large.tar.gz.
['./notMNIST_large/A', './notMNIST_large/B', './notMNIST_large/C', './notMNIST_large/D', './notMNIST_large/E', './notMNIST_large/F', './notMNIST_large/G', './notMNIST_large/H', './notMNIST_large/I', './notMNIST_large/J']
./notMNIST_small already present - Skipping extraction of ./notMNIST_small.tar.gz.
['./notMNIST_small/A', './notMNIST_small/B', './notMNIST_small/C', './notMNIST_small/D', './notMNIST_small/E', './notMNIST_small/F', './notMNIST_small/G', './notMNIST_small/H', './notMNIST_small/I', './notMNIST_small/J']
3. 加载数据
验证数据集，查看一下A目录里面前20个数据图片
fn = os.listdir("notMNIST_small/A/")
for file in fn[:20]:
path = 'notMNIST_small/A/' + file
display(Image(path))
现在我们将要把图片数据转换成为像素，并且对数据进行Zero Mean是数据更加正则化，整个数据集会被加载到一个3D数组中（图片index，x，y）。如果有些图片不能读取，我们就直接忽略掉。
由于可能不能一次性将所有数据读取到内存中，我们会对每个目录图片分别处理，并将处理完的数据存储到对应的pickle文件中。
image_size = 28 # Pixel width and height.
pixel_depth = 255.0 # Number of levels per pixel. def load_letter(folder, min_num_images):
"""Load the data for a single letter label."""
image_files = os.listdir(folder)
dataset = np.ndarray(shape=(len(image_files), image_size, image_size),
dtype=np.float32)
print(folder)
num_images = 0
for image in image_files:
image_file = os.path.join(folder, image)
try:
image_data = (ndimage.imread(image_file).astype(float) -
pixel_depth / 2) / pixel_depth
if image_data.shape != (image_size, image_size):
raise Exception('Unexpected image shape: %s' % str(image_data.shape))
dataset[num_images, :, :] = image_data
num_images = num_images + 1
except IOError as e:
print('Could not read:', image_file, ':', e, '- it\'s ok, skipping.') dataset = dataset[0:num_images, :, :]
if num_images < min_num_images:
raise Exception('Many fewer images than expected: %d < %d' %
(num_images, min_num_images)) print('Full dataset tensor:', dataset.shape)
print('Mean:', np.mean(dataset))
print('Standard deviation:', np.std(dataset))
return dataset def maybe_pickle(data_folders, min_num_images_per_class, force=False):
dataset_names = []
for folder in data_folders:
set_filename = folder + '.pickle'
dataset_names.append(set_filename)
if os.path.exists(set_filename) and not force:
# You may override by setting force=True.
print('%s already present - Skipping pickling.' % set_filename)
else:
print('Pickling %s.' % set_filename)
dataset = load_letter(folder, min_num_images_per_class)
try:
with open(set_filename, 'wb') as f:
pickle.dump(dataset, f, pickle.HIGHEST_PROTOCOL)
except Exception as e:
print('Unable to save data to', set_filename, ':', e) return dataset_names train_datasets = maybe_pickle(train_folders, 45000)
test_datasets = maybe_pickle(test_folders, 1800)
输入如下：
notMNIST_large/A.pickle already present - Skipping pickling.
notMNIST_large/B.pickle already present - Skipping pickling.
notMNIST_large/C.pickle already present - Skipping pickling.
notMNIST_large/D.pickle already present - Skipping pickling.
notMNIST_large/E.pickle already present - Skipping pickling.
notMNIST_large/F.pickle already present - Skipping pickling.
notMNIST_large/G.pickle already present - Skipping pickling.
notMNIST_large/H.pickle already present - Skipping pickling.
notMNIST_large/I.pickle already present - Skipping pickling.
notMNIST_large/J.pickle already present - Skipping pickling.
notMNIST_small/A.pickle already present - Skipping pickling.
notMNIST_small/B.pickle already present - Skipping pickling.
notMNIST_small/C.pickle already present - Skipping pickling.
notMNIST_small/D.pickle already present - Skipping pickling.
notMNIST_small/E.pickle already present - Skipping pickling.
notMNIST_small/F.pickle already present - Skipping pickling.
notMNIST_small/G.pickle already present - Skipping pickling.
notMNIST_small/H.pickle already present - Skipping pickling.
notMNIST_small/I.pickle already present - Skipping pickling.
notMNIST_small/J.pickle already present - Skipping pickling. 验证数据，我们从A.pickle中随机挑选了一个数据
# index 0 should be all As, 1 = all Bs, etc.
pickle_file = train_datasets[0] # With would automatically close the file after the nested block of code
with open(pickle_file, 'rb') as f: # unpickle
letter_set = pickle.load(f) # pick a random image index
sample_idx = np.random.randint(len(letter_set)) # extract a 2D slice
sample_image = letter_set[sample_idx, :, :]
plt.figure() # display it
plt.imshow(sample_image)
4. 准备训练数据、验证数据和测试数据
我们将pickle文件读取出来进行合并，生成了对应的训练数据（Training）,验证数据（Validation）和测试数据（Testing）。
def make_arrays(nb_rows, img_size):
if nb_rows:
dataset = np.ndarray((nb_rows, img_size, img_size), dtype=np.float32)
labels = np.ndarray(nb_rows, dtype=np.int32)
else:
dataset, labels = None, None
return dataset, labels def merge_datasets(pickle_files, train_size, valid_size=0):
num_classes = len(pickle_files)
valid_dataset, valid_labels = make_arrays(valid_size, image_size)
train_dataset, train_labels = make_arrays(train_size, image_size)
vsize_per_class = valid_size // num_classes
tsize_per_class = train_size // num_classes start_v, start_t = 0, 0
end_v, end_t = vsize_per_class, tsize_per_class
end_l = vsize_per_class+tsize_per_class
for label, pickle_file in enumerate(pickle_files):
try:
with open(pickle_file, 'rb') as f:
letter_set = pickle.load(f)
# let's shuffle the letters to have random validation and training set
np.random.shuffle(letter_set)
if valid_dataset is not None:
valid_letter = letter_set[:vsize_per_class, :, :]
valid_dataset[start_v:end_v, :, :] = valid_letter
valid_labels[start_v:end_v] = label
start_v += vsize_per_class
end_v += vsize_per_class train_letter = letter_set[vsize_per_class:end_l, :, :]
train_dataset[start_t:end_t, :, :] = train_letter
train_labels[start_t:end_t] = label
start_t += tsize_per_class
end_t += tsize_per_class
except Exception as e:
print('Unable to process data from', pickle_file, ':', e)
raise return valid_dataset, valid_labels, train_dataset, train_labels train_size = 200000
valid_size = 10000
test_size = 10000 valid_dataset, valid_labels, train_dataset, train_labels = merge_datasets(
train_datasets, train_size, valid_size)
_, _, test_dataset, test_labels = merge_datasets(test_datasets, test_size) print('Training:', train_dataset.shape, train_labels.shape)
print('Validation:', valid_dataset.shape, valid_labels.shape)
print('Testing:', test_dataset.shape, test_labels.shape)
Training: (200000, 28, 28) (200000,)
Validation: (10000, 28, 28) (10000,)
Testing: (10000, 28, 28) (10000,)
随后将数据随机排列
def randomize(dataset, labels):
permutation = np.random.permutation(labels.shape[0])
shuffled_dataset = dataset[permutation,:,:]
shuffled_labels = labels[permutation]
return shuffled_dataset, shuffled_labels
train_dataset, train_labels = randomize(train_dataset, train_labels)
test_dataset, test_labels = randomize(test_dataset, test_labels)
valid_dataset, valid_labels = randomize(valid_dataset, valid_labels)
将数据保存到notMNIST.pickle文件
pickle_file = 'notMNIST.pickle' try:
f = open(pickle_file, 'wb')
save = {
'train_dataset': train_dataset,
'train_labels': train_labels,
'valid_dataset': valid_dataset,
'valid_labels': valid_labels,
'test_dataset': test_dataset,
'test_labels': test_labels,
}
pickle.dump(save, f, pickle.HIGHEST_PROTOCOL)
f.close()
except Exception as e:
print('Unable to save data to', pickle_file, ':', e)
raise
去除数据集中重复的数据
import time def check_overlaps(images1, images2):
images1.flags.writeable=False
images2.flags.writeable=False
start = time.clock()
hash1 = set([hash(image1.data) for image1 in images1])
hash2 = set([hash(image2.data) for image2 in images2])
all_overlaps = set.intersection(hash1, hash2)
return all_overlaps, time.clock()-start r, execTime = check_overlaps(train_dataset, test_dataset)
print('Number of overlaps between training and test sets: {}. Execution time: {}.'.format(len(r), execTime)) r, execTime = check_overlaps(train_dataset, valid_dataset)
print('Number of overlaps between training and validation sets: {}. Execution time: {}.'.format(len(r), execTime)) r, execTime = check_overlaps(valid_dataset, test_dataset)
print('Number of overlaps between validation and test sets: {}. Execution time: {}.'.format(len(r), execTime))
Number of overlaps between training and test sets: 1153. Execution time: 0.951144.
Number of overlaps between training and validation sets: 952. Execution time: 1.014579.
Number of overlaps between validation and test sets: 55. Execution time: 0.088879.
5. 训练模型
我们使用逻辑回归模型来进行训练，来看看最终的准确度如何？
samples, width, height = train_dataset.shape
X_train = np.reshape(train_dataset,(samples,width*height))
y_train = train_labels # Prepare testing data
samples, width, height = test_dataset.shape
X_test = np.reshape(test_dataset,(samples,width*height))
y_test = test_labels # Import
from sklearn.linear_model import LogisticRegression # Instantiate
lg = LogisticRegression(multi_class='multinomial', solver='lbfgs', random_state=42, verbose=1, max_iter=1000, n_jobs=-1) # Fit
lg.fit(X_train, y_train) # Predict
y_pred = lg.predict(X_test) # Score
from sklearn import metrics
metrics.accuracy_score(y_test, y_pred)
[Parallel(n_jobs=-1)]: Done 1 out of 1 | elapsed: 4.6min finished
0.90059999999999996
大概花费了5分钟的时间，训练出来的回归模型准确率达到了90%，不错的尝试了！