def gradient(self, network, x_vec, y_vec):
# ランダムにバッチを取得
batch_mask = np.random.choice(len(x_vec), self.batch_size)
# ミニバッチに対応する教師訓練ラベルデータを取得
x_batch = x_vec[batch_mask]
# ミニバッチに対応する訓練正解ラベルデータを取得する
y_batch = y_vec[batch_mask]
# y_batch = y_batch[:, np.newaxis]
z1, z2, y = self.forward(network, x_batch)
grad = self.backward(x_batch, y_batch, z1, z2, y)
# optimizerの設定 モメンタムを利用
opt = optimizer.Adam(self.learning_rate)
opt.update(network, grad)
return y_batch, y
'filter_num': 10,
'filter_size': 7,
'pad': 1,
'stride': 1
},
conv_param_2={
'filter_num': 20,
'filter_size': 3,
'pad': 1,
'stride': 1
},
hidden_size=100,
output_size=10,
weight_init_std=0.01)
optimizer = optimizer.Adam()
# 時間がかかるため100に設定
iters_num = 100
# iters_num = 1000
train_size = x_train.shape[0]
batch_size = 100
train_loss_list = []
accuracies_train = []
accuracies_test = []
plot_interval = 10
for i in range(iters_num):
batch_mask = np.random.choice(train_size, batch_size)
correct += check
return correct/len(data)
def run(model, train_set, vali_set, test_set):
for epoch in range(1, 300):
train_loss = train(model, train_set)
vali_loss = validation(model, vali_set)
accuracy = test(model, test_set)
print("epoch:", epoch, "\ttrain_loss:", train_loss, "\tvali_loss:", vali_loss, "\taccuracy:", accuracy)
lr = 0.01
model = net.model(optimizer.Adam(lr=lr)) # 30 66
#model = net.model(optimizer.GradientDescent(lr=lr)) #30번에 32퍼 학,검,테 데이터셋 128개일때
model.add(nn.conv2d(filters=32, kernel_size=[3,3], strides=[1,1], w_init=init.he))
model.add(nn.relu())
model.add(nn.maxpool2d(kernel_size=[2,2], strides=[2,2]))
model.add(nn.dropout(0.6))
model.add(nn.conv2d(filters=64, kernel_size=[3,3], strides=[1,1], w_init=init.he))
model.add(nn.relu())
model.add(nn.maxpool2d(kernel_size=[2,2], strides=[2,2]))
model.add(nn.dropout(0.6))
model.add(nn.conv2d(filters=128, kernel_size=[3,3], strides=[1,1], w_init=init.he))
model.add(nn.relu())
model.add(nn.maxpool2d(kernel_size=[2,2], strides=[2,2]))
},
hidden_size_1=120,
hidden_size_2=84,
output_size=10,
weight_init_std=0.01)
# for layer in network.layers.values():
# print(layer)
# print(network.lossLayer)
# print('****** Print structure without values: OK ******')
train_x_batch, train_y_batch = get_one_batch(train_x,
train_y,
batch_size=10)
show_structure(network, train_x_batch, train_y_batch)
op = optimizer.Adam(lr=0.001)
epoch = 100
for i in range(5000):
train_x_batch, train_y_batch = get_one_batch(train_x,
train_y,
batch_size=30)
grads = network.gradient(train_x_batch, train_y_batch)
try:
op.update(network.params, grads)
except ZeroDivisionError as e:
print('Handling run-time error:', e)
if i % epoch == 0:
# calculate accuracy
train_acc = network.accuracy(train_x_batch, train_y_batch)
train_acc_list.append(train_acc)
