Ex 1: Visualization of MLP weights on MNIST
此範例將使用MNIST dataset的訓練資料集去訓練MLPClassifier，資料集中每張圖片都是28*28，對於第一層的每個神經元都會有28*28個特徵，輸出結果是將訓練資料的每個像素點對於神經元的權重畫成28*28的圖，用來表示圖片上每個像素點對於神經元的權重多寡。
1.matplotlib.pyplot:用來繪製影像 2.sklearn.datasets:引入內建的手寫數字資料庫 3.sklearn.neural_network:引入類神經網路的套件
import matplotlib.pyplot as pltfrom sklearn.datasets import fetch_mldatafrom sklearn.neural_network import MLPClassifiermnist = fetch_mldata("MNIST original")
# 將灰階影像降尺度降到[0,1]X, y = mnist.data / 255., mnist.targetX_train, X_test = X[:60000], X[60000:]y_train, y_test = y[:60000], y[60000:]
#hidden_layer_sizes=(50)此處使用1層隱藏層，只有50個神經元，max_iter=10疊代訓練10次mlp = MLPClassifier(hidden_layer_sizes=(50), max_iter=10, alpha=1e-4,solver='sgd', verbose=10, tol=1e-4, random_state=1,learning_rate_init=.1)mlp.fit(X_train, y_train)#畫出16個神經元的權重圖，黑色表示負的權重，越深色表示數值越大，白色表示正的權重，越淺色表示數值越大fig, axes = plt.subplots(4, 4)# use global min / max to ensure all weights are shown on the same scalevmin, vmax = mlp.coefs_[0].min(), mlp.coefs_[0].max()for coef, ax in zip(mlp.coefs_[0].T, axes.ravel()):ax.matshow(coef.reshape(28, 28), cmap=plt.cm.gray, vmin=.5 * vmin,vmax=.5 * vmax)ax.set_xticks(())ax.set_yticks(())plt.show()
圖1:16個神經元對於影像的權重圖
圖2:疊代計算時loss下降
print(__doc__)import matplotlib.pyplot as pltfrom sklearn.datasets import fetch_mldatafrom sklearn.neural_network import MLPClassifiermnist = fetch_mldata("MNIST original")X, y = mnist.data / 255., mnist.targetX_train, X_test = X[:60000], X[60000:]y_train, y_test = y[:60000], y[60000:]mlp = MLPClassifier(hidden_layer_sizes=(50), max_iter=10, alpha=1e-4,solver='sgd', verbose=10, tol=1e-4, random_state=1,learning_rate_init=.1)mlp.fit(X_train, y_train)print("Training set score: %f" % mlp.score(X_train, y_train))print("Test set score: %f" % mlp.score(X_test, y_test))fig, axes = plt.subplots(4, 4)vmin, vmax = mlp.coefs_[0].min(), mlp.coefs_[0].max()for coef, ax in zip(mlp.coefs_[0].T, axes.ravel()):ax.matshow(coef.reshape(28, 28), cmap=plt.cm.gray, vmin=.5 * vmin,vmax=.5 * vmax)ax.set_xticks(())ax.set_yticks(())plt.show()