这是Keras Sequential模型学习吗？

Question

我正在尝试构建一个适合我的数据的Keras Sequential模型。然而，我在选择图层和设置输入形状时遇到了麻烦。我的模型精度从0.4943开始，并且不会在不同时期之间发生变化。看起来我的模型不是在学习。

数据如下所示：

X = [[[0.00000000e+00 0.00000000e+00]
  [1.82562794e-01 6.81775296e-01]
  [1.13191538e+00 1.37766573e+00]
  ...
  [5.31509230e+01 4.88222520e+01]
  [5.38463488e+01 4.92077884e+01]
  [5.43891348e+01 4.98190918e+01]]

 [[0.00000000e+00 0.00000000e+00]
  [4.81657107e-01 4.62969773e-01]
  [1.33733394e+00 8.20860280e-01]
  ...
  [5.00154741e+01 4.49145568e+01]
  [5.06145436e+01 4.58551323e+01]
  [5.14753045e+01 4.66484598e+01]]

 [[0.00000000e+00 0.00000000e+00]
  [1.24209617e-01 3.41455813e-01]
  [6.62306377e-01 9.70226310e-01]
  ...
  [4.59534909e+01 5.14811676e+01]
  [4.65830639e+01 5.15458682e+01]
  [4.69169909e+01 5.18978055e+01]]

 ...

 [[0.00000000e+00 0.00000000e+00]
  [8.37513698e-01 2.36545136e-01]
  [2.09606414e+00 2.18579855e+00]
  ...
  [9.33516241e+01 9.02639438e+01]
  [9.48198248e+01 9.09696034e+01]
  [9.56924057e+01 9.11994364e+01]]

 [[0.00000000e+00 0.00000000e+00]
  [1.16628793e+00 3.07939104e-01]
  [2.90856042e+00 1.93300849e+00]
  ...
  [9.50615310e+01 9.54437621e+01]
  [9.64466547e+01 9.62387560e+01]
  [9.84132452e+01 9.68517902e+01]]

 [[0.00000000e+00 0.00000000e+00]
  [7.07518408e-02 1.63762559e+00]
  [1.47380576e+00 3.01519861e+00]
  ...
  [9.56341427e+01 8.22719298e+01]
  [9.75264435e+01 8.41242858e+01]
  [9.85001877e+01 8.44169342e+01]]]

X.shape = (2000, 100, 2)

y = [0. 0. 0. ... 1. 1. 1.]

y.shape = (2000,)

下面是模型代码：

model = keras.Sequential([
    keras.layers.Flatten(input_shape=(100,2)),
    keras.layers.Dense(16, activation=tf.nn.relu),
    keras.layers.Dense(16, activation=tf.nn.relu),
    keras.layers.Dense(1, activation=tf.nn.sigmoid),
])

model.compile(optimizer='adam',
              loss='binary_crossentropy',
              metrics=['accuracy'])

h = model.fit(X_train, y_train, epochs=50, batch_size=3, shuffle=True)

test_loss, test_acc = model.evaluate(X_test, y_test)
print('Test accuracy:', test_acc)

我正在试着做一个二进制分类。我的模型有问题吗？任何帮助都是非常感谢的。

Answer 1

我注意到你的batch_size=3不是一个好的选择。尝试使用不同的架构/参数进行实验。下面是一个简单的例子：

model = Sequential()
# model.add(Dense(256, input_dim=2, activation='relu'))
model.add(Dense(256, input_shape=X.shape, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu')
model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam',
              loss='binary_crossentropy',
               metrics=['accuracy'])

model.fit(X_train, y_train, epochs=10, batch_size=64, shuffle=True)

工作示例：

from keras import Sequential
from keras.layers import Dense, Dropout

# sample data, 100 samples
X_train = np.random.random((100,2)) 
y_train = np.random.randint(2, size=(100, 1)) 

model = Sequential()
model.add(Dense(256, input_dim=2, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam',
              loss='binary_crossentropy',
               metrics=['accuracy'])

model.fit(X_train, y_train, epochs=10, batch_size=2)
model.predict(X_train).shape # (100, 1)

Answer 2

在解决了这个问题之后，我意识到将输入形状更改为(200，)可以解决这个问题。当您向数据添加维度时，事情会变得复杂。谢谢你@YOLO帮助我。