theano學習之分類學習
阿新 • • 發佈:2018-12-28
程式碼:
from __future__ import print_function import numpy as np import theano import theano.tensor as T #該函式功能用來計算準確率 def compute_accuracy(y_target, y_predict): correct_prediction = np.equal(y_predict, y_target) accuracy = np.sum(correct_prediction)/len(correct_prediction) return accuracy # generate a dataset: D = (input_values, target_class) #用 randn 隨機生成資料集。 D 中的 input_values 是 400 個樣本,784 個feature。 # target_class 是有兩類,0 和 1。 要做的是,用神經網路訓練資料學習哪些輸入對應 0,哪些對應1 # 生成隨機數: D = (input_values, target_class) rng = np.random N = 400 # training sample size feats = 784 # number of input variables D = (rng.randn(N, feats), rng.randint(size=N, low=0, high=2))#D[0]表特徵,D[1]表標籤 print(D[0]) print(D[1]) #定義x,y容器,相當於TensorFlow中的placeholder,用來存資料 x = T.dmatrix("x") y = T.dvector("y") #初始化權重和偏置 W = theano.shared(rng.randn(feats), name="w") b = theano.shared(0., name="b") #定義啟用函式和交叉熵 p_1 = T.nnet.sigmoid(T.dot(x, W) + b) #啟用函式 prediction = p_1 > 0.5 # The prediction thresholded閾值,呼叫了p_1函式,預測當前權重下的值,這個是個布林值,ture或false xent = -y * T.log(p_1) - (1-y) * T.log(1-p_1) # 交叉熵 # or # xent = T.nnet.binary_crossentropy(p_1, y) # this is provided by theano cost = xent.mean() + 0.01 * (W ** 2).sum()#正則化後的代價函式(L2正則化) gW, gb = T.grad(cost, [W, b]) # 梯度 #啟用網路 learning_rate = 0.1 train = theano.function( inputs=[x, y], outputs=[prediction, xent.mean()], updates=((W, W - learning_rate * gW), (b, b - learning_rate * gb))) predict = theano.function(inputs=[x], outputs=prediction) #訓練模型 for i in range(500): pred, err = train(D[0], D[1])#D[0]指特徵所有資料,D[1]指標籤所有資料 if i % 50 == 0: print('cost:', err) print("accuracy:", compute_accuracy(D[1], predict(D[0]))) print("target values for D:") print(D[1]) print("prediction on D:") print(predict(D[0]))