TFlearn到范畴

Question

我正在使用tflearn的DNN，我想改变我的特征和标签，使之成为绝对的，而不是数字的。

这是我的网：

x = tf.placeholder(dtype= tf.float32, shape=[None, 6], name='x')
# Build neural network
input_layer = tflearn.input_data(shape=[None, 6])
net = input_layer
net = tflearn.fully_connected(net, 128, activation='relu')
net = tflearn.fully_connected(net, 64, activation='relu')
net = tflearn.fully_connected(net, 16, activation='relu')
net = tflearn.fully_connected(net, 2, activation='sigmoid')
net = tflearn.regression(net, optimizer='adam', loss='mean_square', metric='R2')

w = tf.Variable(tf.truncated_normal([2, 2], stddev=0.1))
b = tf.Variable(tf.constant(1.0, shape=[2]))
y = tf.nn.softmax(tf.matmul(net, w) + b, name='y')

model = tflearn.DNN(net, tensorboard_verbose=3)
return model

我知道tflearn.data_utils.to_categorical，但我不知道如何注入这种方法。谢谢

编辑：--我尝试了一些东西，比如：

train_goal = tflearn.data_utils.to_categorical(train_goal, nb_classes=2)
            test_goal = tflearn.data_utils.to_categorical(test_goal, nb_classes=2)

也改变了损失：

net = tflearn.regression(net, optimizer='adadelta',  loss='categorical_crossentropy', metric= self.accuracy)

但我的损失超过了1：

Training Step: 35  | total loss: 1.64734 | time: 1.322s
| AdaDelta | epoch: 001 | loss: 1.64734 - acc: 1.0000 | val_loss: 1.64313 - val_acc: 1.0000 -- iter: 2204/2204
--
Training Step: 70  | total loss: 1.61961 | time: 0.216s
| AdaDelta | epoch: 002 | loss: 1.61961 - acc: 1.0000 | val_loss: 0.00000 - val_acc: 0.0000 -- iter: 2204/2204
--
Training Step: 105  | total loss: 1.58511 | time: 1.188s
| AdaDelta | epoch: 003 | loss: 1.58511 - acc: 1.0000 | val_loss: 1.57300 - val_acc: 1.0000 -- iter: 2204/2204

问题出在哪里？

Answer 1

我有一个类似的错误，也很高的损失。尝试使用train_goal.T而不是train_goal。确保to_categorical的输入y具有类似于(n，)的形状