def plot_nmf_faces():
"""利用越来越多的非负分量对三张人脸图像进行重建"""
# 下面这张图计算时间比较长，需要耐心等等
people = load_people()
image_shape = people.images[0].shape
X_train, X_test, y_train, y_test = load_train_test_faces()
mglearn.plots.plot_nmf_faces(X_train, X_test, image_shape=image_shape)
plt.suptitle("图3-14 利用越来越多的非负分量对三张人脸图像进行重建")
def knn_classify_pca_faces():
people = load_people()
X_train, X_test, y_train, y_test = load_train_test_faces()
# 计算每个目标出现的次数
counts = np.bincount(people.target)
# 将次数与目标名称一起打印出来
print('{0:25} {1:5}'.format("姓名", "照片数目"))
for i, (count, name) in enumerate(zip(counts, people.target_names)):
print("{0:25} {1:3}".format(name, count), end='\t\t')
if (i + 1) % 3 == 0:
print()
print()
# 2) 使用KNN训练和测试被PCA白化的数据
# X_train_pca.shape: (1341, 100)
# Test set score of 1-nn: 0.35
# pca.components_.shape: (100, 5655)
from sklearn.decomposition import PCA
pca = PCA(n_components=75, whiten=True, random_state=seed)
pca.fit(X_train)
X_train_pca = pca.transform(X_train)
X_test_pca = pca.transform(X_test)
from sklearn.neighbors import KNeighborsClassifier
knn = KNeighborsClassifier(n_neighbors=1)
knn.fit(X_train_pca, y_train)
print('=' * 20)
print("-- 使用KNN训练和测试经过PCA白化的数据 --")
print('PCA主成分的形状: {}'.format(pca.components_.shape))
print('经过PCA白化的数据的形状: {}'.format(X_train_pca.shape))
print('PCA白化的数据经过KNN训练后测试集的精度: {:.2f}'.format(knn.score(X_test_pca, y_test)))
image_shape = people.images[0].shape
fig, axes = plt.subplots(3, 5, figsize=(20, 10), subplot_kw={'xticks': (), 'yticks': ()})
for i, (component, ax) in enumerate(zip(pca.components_, axes.ravel())):
ax.imshow(component.reshape(image_shape), cmap='viridis')
ax.set_title('{}.component'.format((i + 1)))
pass
plt.suptitle("图3-9：人脸数据集的前15个主成分的成分向量")
# 图3-10：人脸照片=Σ_(i=0)^(n) x_i * components_i
# 每张照片就是主成分的加权求和
mglearn.plots.plot_pca_faces(X_train, X_test, image_shape)
plt.suptitle("图3-11 利用越来越多的主成分对三张人脸图像进行重建")
# plt.figure()
# plt.scatter(X_train_pca[:, 0], X_train_pca[:, 1], y_train)
plt.figure()
mglearn.discrete_scatter(X_train_pca[:, 0], X_train_pca[:, 1], y_train)
plt.xlabel('第一个主成分')
plt.ylabel('第二个主成分')
plt.suptitle("两个主成分的散点图\n观察数据聚类的效果(数据不可分)")
def knn_classify_nmf_faces():
people = load_people()
image_shape = people.images[0].shape
X_train, X_test, y_train, y_test = load_train_test_faces()
from sklearn.neighbors import KNeighborsClassifier
knn = KNeighborsClassifier(n_neighbors=1)
knn.fit(X_train, y_train)
print('=' * 20)
print("-- 使用KNN训练和测试原始数据 --")
print('原始数据的形状: {}'.format(X_train.shape))
print('原始数据经过KNN训练后测试集的精度: {:.2f}'.format(knn.score(X_test, y_test)))
# 分量太少，学习的精确度较差，但是增加分量不一定能够提高模型的精度，但是会增加计算时间
from sklearn.decomposition import NMF
nmf = NMF(n_components=100, max_iter=200, random_state=seed)
nmf.fit(X_train)
X_train_nmf = nmf.transform(X_train)
X_test_nmf = nmf.transform(X_test)
from sklearn.neighbors import KNeighborsClassifier
knn = KNeighborsClassifier(n_neighbors=1)
knn.fit(X_train_nmf, y_train)
print('=' * 20)
print("-- 使用KNN训练和测试经过NMF处理的数据 --")
print('NMF成分的形状: {}'.format(nmf.components_.shape))
print('经过NMF的数据的形状: {}'.format(X_train_nmf.shape))
print('NMF的数据经过KNN训练后测试集的精度: {:.2f}'.format(knn.score(X_test_nmf, y_test)))
fig, axes = plt.subplots(3, 5, figsize=(20, 10), subplot_kw={'xticks': (), 'yticks': ()})
plt.suptitle("图3-15 使用15个分量的NMF在人脸数据集上找到的15个分量")
for i, (component, ax) in enumerate(zip(nmf.components_, axes.ravel())):
ax.imshow(component.reshape(image_shape), cmap='viridis')
ax.set_title('{}.component'.format((i + 1)))
pass
components = 3 # 不同分量的图片有一定的共性
indexes = np.argsort(X_train_nmf[:, components])[::-1]
fig, axes = plt.subplots(2, 5, figsize=(20, 10), subplot_kw={'xticks': (), 'yticks': ()})
plt.suptitle("图3-16 第3个分量的系数较大的人脸")
for i, (index, ax) in enumerate(zip(indexes, axes.ravel())):
ax.imshow(X_train[index].reshape(image_shape))
pass
components = 7 # 不同分量的图片有一定的共性
indexes = np.argsort(X_train_nmf[:, components])[::-1]
fig, axes = plt.subplots(2, 5, figsize=(20, 10), subplot_kw={'xticks': (), 'yticks': ()})
plt.suptitle("图3-16 第7个分量的系数较大的人脸")
for i, (index, ax) in enumerate(zip(indexes, axes.ravel())):
ax.imshow(X_train[index].reshape(image_shape))
pass
def knn_classify_original_faces():
X_train, X_test, y_train, y_test= load_train_test_faces()
# 1) 使用KNN训练和测试数据
# 5655=87*65
# X_train.shape: (1341, 5655)
# Test set score of 1-knn: 0.27
from sklearn.neighbors import KNeighborsClassifier
knn = KNeighborsClassifier(n_neighbors=1)
knn.fit(X_train, y_train)
print('=' * 20)
print("-- 使用KNN训练和测试原始数据 --")
print('原始数据的形状: {}'.format(X_train.shape))
print('原始数据经过KNN训练后测试集的精度: {:.2f}'.format(knn.score(X_test, y_test)))
def kmeans_vector_quantization():
people = load_people()
image_shape = people.images[0].shape
X_train, X_test, y_train, y_test = load_train_test_faces()
from sklearn.cluster import KMeans
kmeans = KMeans(n_clusters=100, random_state=seed)
kmeans.fit(X_train)
from sklearn.decomposition import PCA
pca = PCA(n_components=100, random_state=seed)
pca.fit(X_train)
from sklearn.decomposition import NMF
nmf = NMF(n_components=100, random_state=seed)
nmf.fit(X_train)
fig, axes = plt.subplots(3,
15,
figsize=(20, 10),
subplot_kw={
'xticks': (),
'yticks': ()
})
fig.suptitle('图3-30：对比K均值的簇中心与PCA和NMF找到的分量')
# K均值找到的是图片的共性，PCA 找到的是图片变化最大的特征，NMF 找到的是图片中基础元素
for ax, comp_kmeans, comp_pca, comp_nmf in zip(axes.T,
kmeans.cluster_centers_,
pca.components_,
nmf.components_):
ax[0].imshow(comp_kmeans.reshape(image_shape))
ax[1].imshow(comp_pca.reshape(image_shape), cmap='viridis')
ax[2].imshow(comp_nmf.reshape(image_shape))
axes[0, 0].set_ylabel('kmeans')
axes[1, 0].set_ylabel('pca')
axes[2, 0].set_ylabel('nmf')
X_reconstructed_kmeans = kmeans.cluster_centers_[kmeans.predict(X_test)]
X_reconstructed_pca = pca.inverse_transform(pca.transform(X_test))
X_reconstructed_nmf = np.dot(nmf.transform(X_test), nmf.components_)
fig, axes = plt.subplots(4,
5,
figsize=(20, 10),
subplot_kw={
'xticks': (),
'yticks': ()
})
fig.suptitle('图3-31：利用100个分量（或簇中心）的K均值、PCA和NMF的图像重建的对比——K均值的每张图像仅使用了一个簇中心')
# K均值重建的效果没有其他两种好，因为K均值取的是平均值，而每个类别有自己的特点，而不是所有特点的平均
for ax, orig, rec_kmeans, rec_pca, rec_nmf in zip(axes.T, X_test,
X_reconstructed_kmeans,
X_reconstructed_pca,
X_reconstructed_nmf):
ax[0].imshow(orig.reshape(image_shape))
ax[1].imshow(rec_kmeans.reshape(image_shape))
ax[2].imshow(rec_pca.reshape(image_shape))
ax[3].imshow(rec_nmf.reshape(image_shape))
axes[0, 0].set_ylabel('original')
axes[1, 0].set_ylabel('kmeans')
axes[2, 0].set_ylabel('pca')
axes[3, 0].set_ylabel('nmf')