我构建了一个简单的Keras网络：
import numpy as np;from keras.models import Sequential;from keras.layers import Dense,Activation;data= np.genfromtxt("./kerastests/mydata.csv", delimiter=';')x_target=data[:,29]x_training=np.delete(data,6,axis=1)x_training=np.delete(x_training,28,axis=1)model=Sequential()model.add(Dense(20,activation='relu', input_dim=x_training.shape[1]))model.add(Dense(10,activation='relu'))model.add(Dense(1));model.compile(optimizer='adam',loss='mean_squared_error',metrics=['accuracy'])model.fit(x_training, x_target)从我的源数据中，我删除了两列，如你所见。一列是包含字符串格式日期的列（在数据集中，除了这一列外，我还有一列是天，一列是月，一列是年，所以我不需要那一列），另一列是我用作模型目标的列。
当我训练这个模型时，我得到了这样的输出：
32/816 [>.............................] - ETA: 23s - loss: 13541942.0000 - acc: 0.0000e+00800/816 [============================>.] - ETA: 0s - loss: 11575466.0400 - acc: 0.0000e+00 816/816 [==============================] - 1s - loss: 11536905.2353 - acc: 0.0000e+00 Epoch 2/10 32/816 [>.............................] - ETA: 0s - loss: 6794785.0000 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5381360.4314 - acc: 0.0000e+00 Epoch 3/10 32/816 [>.............................] - ETA: 0s - loss: 6235184.0000 - acc: 0.0000e+00800/816 [============================>.] - ETA: 0s - loss: 5199512.8700 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5192977.4216 - acc: 0.0000e+00 Epoch 4/10 32/816 [>.............................] - ETA: 0s - loss: 4680165.5000 - acc: 0.0000e+00736/816 [==========================>...] - ETA: 0s - loss: 5050110.3043 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5168771.5490 - acc: 0.0000e+00 Epoch 5/10 32/816 [>.............................] - ETA: 0s - loss: 5932391.0000 - acc: 0.0000e+00768/816 [===========================>..] - ETA: 0s - loss: 5198882.9167 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5159585.9020 - acc: 0.0000e+00 Epoch 6/10 32/816 [>.............................] - ETA: 0s - loss: 4488318.0000 - acc: 0.0000e+00768/816 [===========================>..] - ETA: 0s - loss: 5144843.8333 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5151492.1765 - acc: 0.0000e+00 Epoch 7/10 32/816 [>.............................] - ETA: 0s - loss: 6920405.0000 - acc: 0.0000e+00800/816 [============================>.] - ETA: 0s - loss: 5139358.5000 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5169839.2941 - acc: 0.0000e+00 Epoch 8/10 32/816 [>.............................] - ETA: 0s - loss: 3973038.7500 - acc: 0.0000e+00672/816 [=======================>......] - ETA: 0s - loss: 5183285.3690 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5141417.0000 - acc: 0.0000e+00 Epoch 9/10 32/816 [>.............................] - ETA: 0s - loss: 4969548.5000 - acc: 0.0000e+00768/816 [===========================>..] - ETA: 0s - loss: 5126550.1667 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5136524.5098 - acc: 0.0000e+00 Epoch 10/10 32/816 [>.............................] - ETA: 0s - loss: 6334703.5000 - acc: 0.0000e+00768/816 [===========================>..] - ETA: 0s - loss: 5197778.8229 - acc: 0.0000e+00816/816 [==============================] - 0s - loss: 5141391.2059 - acc: 0.0000e+00 为什么会这样？我的数据是一个时间序列。我知道对于时间序列，人们通常不使用Dense神经元，但这只是一个测试。真正困扰我的是准确率总是0。而且，在其他测试中，我甚至遇到了损失值变为”NAN”的情况。
有谁能帮我解决这个问题吗？
回答：
你的模型似乎对应于一个回归模型，原因如下：
-
你在输出层使用了
linear（默认的）作为激活函数（在前一层使用了relu）。 -
你的损失函数是
loss='mean_squared_error'。
然而，你使用的度量标准metrics=['accuracy']适用于分类问题。如果你想做回归，请移除metrics=['accuracy']。也就是说，使用
model.compile(optimizer='adam',loss='mean_squared_error')以下是Keras中用于回归和分类的度量标准列表（摘自这篇博客文章）：
Keras回归度量标准
•均方误差：mean_squared_error, MSE或mse
•平均绝对误差：mean_absolute_error, MAE, mae
•平均绝对百分比误差：mean_absolute_percentage_error, MAPE, mape
•余弦相似度：cosine_proximity, cosine
Keras分类度量标准
•二元准确率：binary_accuracy, acc
•分类准确率：categorical_accuracy, acc
•稀疏分类准确率：sparse_categorical_accuracy
•前k分类准确率：top_k_categorical_accuracy（需要指定k参数）
•稀疏前k分类准确率：sparse_top_k_categorical_accuracy（需要指定k参数）
