def main(config, fileName='data'):
testFile(fileName)
Start = time.time()
msTrain = mnist()
msTest = mnist(kind="t10k")
# Extract Mnist data
trainImg = msTrain.images
trainLabel = msTrain.oneHotLabels
testImg = msTest.images
testLabel = msTest.oneHotLabels
# Training data size
trainSize = trainImg.shape[0]
# Training batch size
batchSize = config['batchSize']
# Number of iterations
itersNum = config['itersNum']
# Learning rate
learningRate = config['learningRate']
# Iter per epoch
iterPerEpoch = max((trainSize / batchSize), 1)
# Initialize bp
network = bpNet(hiddenLayersSize=config['hiddenLayersSize'])
# start training model
trainA = []
testA = []
print("iterPerEpoch:", iterPerEpoch)
#print("Start Model trianing...")
for index in range(itersNum):
if (index % iterPerEpoch == 0):
trainAcc = network.accuracy(trainImg, trainLabel)
testAcc = network.accuracy(testImg, testLabel)
trainA.append(trainAcc)
testA.append(testAcc)
print("Train Acc: %.5f Test Acc: %.5f" % (trainAcc, testAcc))
# Get randomly selected index
trainIndexs = np.random.choice(trainSize, batchSize)
imgs = trainImg[trainIndexs]
labels = trainLabel[trainIndexs]
network.update(imgs, labels, lr=learningRate)
End = time.time()
print("BP Time Consuming:%.2fs" % (End - Start))
# Plot
g = [trainA, testA]
colors = ["blue", "orange"]
labels = ["trainAcc", "testAcc"]
plt.figure(figsize=(6, 6))
for i in range(len(g)):
plt.plot(g[i], colors[i], label=labels[i])
plt.legend(loc=4)
def main(config, fileName='data'):
testFile(fileName)
Start = time.time()
msTrain = mnist()
msTest = mnist(kind="t10k")
# 提取mnist数据
trainImg = msTrain.images
trainLabel = msTrain.oneHotLabels
testImg = msTest.images
testLabel = msTest.oneHotLabels
# 训练数据大小
trainSize = trainImg.shape[0]
# 训练批大小
batchSize = config['batchSize']
# 迭代次数
itersNum = config['itersNum']
# 学习率
learningRate = config['learningRate']
# 轮数
iterPerEpoch = max((trainSize / batchSize), 1)
# bp初始化
network = bpNet(hiddenLayersSize=config['hiddenLayersSize'])
#开始训练模型
trainA = []
testA = []
print("iterPerEpoch:", iterPerEpoch)
#print("Start Model trianing...")
for index in range(itersNum):
if (index % iterPerEpoch == 0):
trainAcc = network.accuracy(trainImg, trainLabel)
testAcc = network.accuracy(testImg, testLabel)
trainA.append(trainAcc)
testA.append(testAcc)
print("训练精度: %.5f 测试精度: %.5f" % (trainAcc, testAcc))
# 获取随机选取的索引
trainIndexs = np.random.choice(trainSize, batchSize)
imgs = trainImg[trainIndexs]
labels = trainLabel[trainIndexs]
network.update(imgs, labels, lr=learningRate)
End = time.time()
print("BP Time Consuming:%.2f秒" % (End - Start))
##作图部分
g = [trainA, testA]
colors = ["blue", "orange"]
labels = ["trainAcc", "testAcc"]
plt.figure(figsize=(6, 6))
for i in range(len(g)):
plt.plot(g[i], colors[i], label=labels[i])
plt.legend(loc=4)
def main():
warnings.filterwarnings("ignore")
print("Start LR Model training...")
msLog = mnist()
msTestLog = mnist(kind="t10k")
Xtrain = msLog.images
Ytrain = msLog.labels
Xtest = msTestLog.images
Ytest = msTestLog.labels
print("Default LR Model loading...")
start = time.time()
model = LogisticRegression(solver="liblinear")
model.fit(Xtrain, Ytrain)
y_pred = model.predict(Xtest)
sum = 0.0
for i in range(10000):
if (y_pred[i] == Ytest[i]):
sum += 1
print('Logistics Test set score: %f' % (sum / 10000.))
end = time.time()
print("Logit Time consuming:%.2fs" % (end - start)) #89.78s
def main():
warnings.filterwarnings("ignore")
print("Start SVM Model loading...")
msSVM = mnist()
msTestSVM = mnist(kind="t10k")
Xtrain = msSVM.images
Ytrain = msSVM.labels
Xtest = msTestSVM.images
Ytest = msTestSVM.labels
# Initialize SVM
print("Default SVM Model loading...")
start = time.time()
clf = svm.SVC()
clf.fit(Xtrain, Ytrain)
predict = clf.predict(Xtest)
score = metrics.accuracy_score(Ytest, predict)
# Test Acc
report = metrics.accuracy_score(Ytest, predict)
end = time.time()
print("SVM Time consuming:%.2f秒" % (end - start))
print(score)
print(report)
def main():
warnings.filterwarnings("ignore")
print("Start RandomForest Model training...")
msLog =mnist()
msTestLog = mnist(kind="t10k")
Xtrain = msLog.images
Ytrain = msLog.labels
Xtest = msTestLog.images
Ytest = msTestLog.labels
# n_estimators=10,max_features=math.sqrt(n_features), max_depth=None,min_samples_split=2, bootstrap=True
print("Default RandomForest model loading...")
start = time.time()
clf = RandomForestClassifier()
clf.fit(Xtrain,Ytrain)
predict = clf.predict(Xtest)
score = metrics.accuracy_score(Ytest,predict)
# Test Accuracy
report = metrics.accuracy_score(Ytest,predict)
print(score)
print(report)
end = time.time()
print("RF Time consuming:%.2f秒"%(end-start))