train_loss_list=[]
train_acc_list=[]
test_acc_list=[]
#1エポック当たりの繰り返し数
iter_per_epoch =max(train_size/batch_size,1)
for i in range(iters_num):
batch_mask = np.random.choice(train_size , batch_size)
x_batch=x_data_train[batch_mask]
t_batch=t_data_train[batch_mask]
#grad=network.numerical_gradient(x_batch,t_batch)
grad=network.gradient(x_batch,t_batch)
for key in("W1","b1","W2","b2"): #Wは大文字
network.params[key] -=learning_rate*grad[key]
loss=network.loss(x_batch,t_batch)
train_loss_list.append(loss)
"""
train_acc = network.accuracy(x_data_train, t_data_train)
test_acc = network.accuracy(x_data_test, t_data_test)
train_acc_list.append(train_acc)
test_acc_list.append(test_acc)
"""
print("ここまで１")
"""
network.predict(引数)で 予測が可能 この引数は入力データの列数と同じであれば行数にかかわらず出てくる
今回は6802やったかなあれを入れたら、onehotの結果が出る
optimizer_mul = AdaGrad()
optimizer_multi = AdaGrad()
for i in range(max_iterations):
batch_mask = np.random.choice(train_size, batch_size)
x_batch = x_train[batch_mask]
t_batch = t_train[batch_mask]
grads_mul = mul_layer_net.gradient(x_batch, t_batch)
grads_multi = multi_layer_net.gradient(x_batch, t_batch)
optimizer_mul.update(mul_layer_net.params, grads_mul)
optimizer_multi.update(multi_layer_net.params, grads_multi)
loss_mul = mul_layer_net.loss(x_batch, t_batch)
loss_multi = multi_layer_net.loss(x_batch, t_batch)
if i % 100 == 0:
print("===========" + "iteration:" + str(i) + "===========")
loss_mul = mul_layer_net.loss(x_batch, t_batch)
loss_multi = multi_layer_net.loss(x_batch, t_batch)
print('mul-loss' + ":" + str(loss_mul))
print('multi-loss' + ":" + str(loss_multi))
# optimizer_function = Adam()
# batch_size = 128
# for i in range(2000):
# batch_index = np.random.choice(x_train.shape[0],batch_size)
# x_batch = x_train[batch_idex]
# t_batch = t_train[batch_index]
# gradient = mul_layer_net.gradient(x_batch, t_batch)
