设置tensorflow顺序模型输入层的形状答案

【问题标题】：Setting the shape of tensorflow sequential model input layer设置tensorflow顺序模型输入层的形状
【发布时间】：2021-11-06 14:16:06
【问题描述】：

我正在尝试为多类分类构建模型，但我不明白如何设置正确的输入形状。我有一个形状为(5420, 212) 的训练集，这是我构建的模型：

model = models.Sequential()
model.add(layers.Dense(64, activation='relu', input_shape = (5420,)))
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(5, activation='softmax'))
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
model.summary()
history = model.fit(X_train, y_train, epochs=20, batch_size=512)

当我运行它时，我得到了错误：

ValueError: Input 0 of layer sequential_9 is incompatible with the layer: expected axis -1 of input shape to have value 5420 but received input with shape (None, 212)

为什么？输入的值不对吗？

【问题讨论】：

你应该设置input_shape = (212,)。
刚试过，我得到了一个 ValueError: Shapes (None, 1) and (None, 5) are incompatible

标签： python tensorflow deep-learning neural-network sequential

【解决方案1】：

输入形状应该等于输入X第二维的长度，而输出形状应该等于输出Y第二维的长度（假设X和Y是二维的，即它们没有更高的维度）。

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from sklearn.datasets import make_classification
from sklearn.preprocessing import OneHotEncoder
tf.random.set_seed(0)

# generate some data
X, y = make_classification(n_classes=5, n_samples=5420, n_features=212, n_informative=212, n_redundant=0, random_state=42)
print(X.shape, y.shape)
# (5420, 212) (5420,)

# one-hot encode the target
Y = OneHotEncoder(sparse=False).fit_transform(y.reshape(-1, 1))
print(X.shape, Y.shape)
# (5420, 212) (5420, 5)

# extract the input and output shapes
input_shape = X.shape[1]   
output_shape = Y.shape[1]   
print(input_shape, output_shape)
# 212 5

# define the model
model = Sequential()
model.add(Dense(64, activation='relu', input_shape=(input_shape,)))
model.add(Dense(64, activation='relu'))
model.add(Dense(output_shape, activation='softmax'))

# compile the model
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

# fit the model
history = model.fit(X, Y, epochs=3, batch_size=512)
# Epoch 1/3
# 11/11 [==============================] - 0s 1ms/step - loss: 4.8206 - accuracy: 0.2208
# Epoch 2/3
# 11/11 [==============================] - 0s 1ms/step - loss: 2.8060 - accuracy: 0.3229
# Epoch 3/3
# 11/11 [==============================] - 0s 1ms/step - loss: 2.0705 - accuracy: 0.3989

【讨论】：