我在使用Keras的函数式API构建一个简单的顺序神经网络。以下是X_train和y_train_encoded（独热编码的y_train，共有10个类别）的形状。

X_train.shape(60000, 28, 28)

y_train_encoded(60000, 10)

我指定了架构，编译并训练如下：

input = keras.layers.Input(shape=(28,28))hidden1 = keras.layers.Dense(128, activation="relu")(input)hidden2 = keras.layers.Dense(128, activation="relu")(hidden1)hidden3 = keras.layers.Dense(28, activation="relu")(hidden2)output = keras.layers.Dense(10, activation="softmax")(hidden3)model = keras.models.Model(inputs=[input], outputs=[output])model.compile(loss="categorical_crossentropy",             optimizer="adam",             metrics=["accuracy"])history=model.fit(X_train, y_train_encoded, epochs=20, validation_split=0.2)

我得到了下面的ValueError错误。

ValueError: Shapes (32, 10) and (32, 28, 10) are incompatible

我想请大家指出我哪里做错了。任何帮助我都会非常感激。

回答：

添加Flatten()层：

input = keras.layers.Input(shape=(28,28))flatten = keras.layers.Flatten()(input)hidden1 = keras.layers.Dense(128, activation="relu")(flatten)hidden2 = keras.layers.Dense(128, activation="relu")(hidden1)hidden3 = keras.layers.Dense(28, activation="relu")(hidden2)output = keras.layers.Dense(10, activation="softmax")(hidden3)model = keras.models.Model(inputs=[input], outputs=[output])