机器学习--周志华--课后习题3.3答案

LR实现，训练集和测试集都是书上给的西瓜数据集3.0α，用keras实现代码如下：

# coding: utf-8"""使用keras实现logistic分类器"""import osimport gzipimport urllibimport numpy as npfrom keras.models import Sequentialfrom keras.layers.core import Dense, Activationfrom keras.optimizers import SGDimport kerasfrom keras.utils.np_utils import to_categoricalfrom keras.callbacks import ModelCheckpoint,Callbackcurrent_dir = os.path.abspath(os.path.curdir)np.random.seed(1337) # 每一次运行结果都一样def load_data():#读入csv数据,每一行的格式都是1,0.697,0.46,是\nfile = open('data/西瓜数据集3.0.csv'.decode('utf-8'))data = [raw.strip('\n').split(',') for raw in file]#X是密度和含糖率两项数据，Y是西瓜是否好瓜的label,1是0否X = [[float(raw[1]), float(raw[2])] for raw in data[1:]]Y = [1 if raw[-1]=='是' else 0 for raw in data[1:]]return X,Ydef create_model(input_dim, output_dim):"""创建logistic模型:param input_dim: (int) 输入维度:param output_dim: (int) 输出维度:return: Sequential"""model = Sequential()model.add(Dense(output_dim, input_dim=input_dim))model.add(Activation('softmax'))sgd = SGD(lr=0.05)model.compile(loss='categorical_crossentropy',optimizer=sgd, metrics=["accuracy"])return modelif __name__ == '__main__':train_set_x, train_set_y = load_data()train_set_x = np.asarray(train_set_x)train_set_y = np.asarray(train_set_y)input_dim = train_set_x.shape[1]output_dim = train_set_y.max() - train_set_y.min() + 1train_set_y = to_categorical(train_set_y)model = create_model(input_dim, output_dim)checkpointer = ModelCheckpoint(filepath="best_model.hdf5", monitor='val_acc', verbose=1, save_best_only=True)# history = LossHistory()earlyStopping = keras.callbacks.EarlyStopping(monitor='val_acc', patience=30, verbose=1, mode='auto')history = model.fit(train_set_x,train_set_y, batch_size=3, nb_epoch=200, callbacks=[checkpointer, earlyStopping],shuffle=True, verbose=0,validation_data=(train_set_x,train_set_y))result = model.predict(train_set_x,batch_size=20,verbose=1)result = [np.argmax(i) for i in result]print resultscore = model.evaluate(train_set_x, train_set_y, verbose=0)print('Test score:', score[0])print('Test accuracy:', score[1])