CNN训练时模型形状不兼容问题：形状匹配错误排查与解决

数据集初始形状 #

print(X_train.shape)
print(X_test.shape)
print(y_train.shape)
print(y_test.shape)
----------------------
(120000, 72)
(12000, 72)
(120000, 6)
(12000, 6)

数据重塑操作 #

X_train = X_train.reshape(len(X_train), X_train.shape[1], 1)
X_test = X_test.reshape(len(X_test), X_test.shape[1], 1)
X_train.shape, X_test.shape
-------------------------------------------------------------------
((120000, 72, 1), (12000, 72, 1))

模型定义 #

def model():
    model = Sequential()
    model.add(Conv1D(filters=64, kernel_size=6, activation='relu', 
                    padding='same', input_shape=(72, 1)))
    model.add(BatchNormalization())
    
    # adding a pooling layer
    model.add(MaxPooling1D(pool_size=(3), strides=2, padding='same'))
    
    model.add(Conv1D(filters=64, kernel_size=6, activation='relu', 
                    padding='same', input_shape=(72, 1)))
    model.add(BatchNormalization())
    model.add(MaxPooling1D(pool_size=(3), strides=2, padding='same'))
    
    model.add(Conv1D(filters=64, kernel_size=6, activation='relu', 
                    padding='same', input_shape=(72, 1)))
    model.add(BatchNormalization())
    model.add(MaxPooling1D(pool_size=(3), strides=2, padding='same'))
    
    model.add(Flatten())
    model.add(Dense(64, activation='relu'))
    model.add(Dense(64, activation='relu'))
    model.add(Dense(3, activation='softmax'))
    
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    return model

报错信息 #

model = model()
model.summary()
logger = CSVLogger('logs.csv', append=True)
his = model.fit(X_train, y_train, epochs=30, batch_size=32, 
          validation_data=(X_test, y_test), callbacks=[logger])

---------------------------------------------------------------
 ValueError: Shapes (32, 6) and (32, 3) are incompatible

问题原因 #

• 模型输出层定义为Dense(3, activation='softmax')，说明模型预期输出3类分类结果，对应形状为(batch_size, 3)
• 但标签数据y_train和y_test的形状是(120000, 6)和(12000, 6)，对应6类分类，两者维度不匹配，导致损失计算时形状冲突

解决方法 #

根据实际分类需求，有两种修正方向：

方向1：实际为6类分类任务 #

修改模型输出层的神经元数量为6，匹配标签维度：

# 将最后一层Dense的3改为6
model.add(Dense(6, activation='softmax'))

方向2：实际为3类分类任务 #

需要将6维标签转换为3维，需结合数据含义调整标签编码：

import numpy as np
import tensorflow as tf

# 示例：假设原6维one-hot标签可按逻辑合并为3类（需根据实际数据调整映射规则）
y_train_idx = np.argmax(y_train, axis=1)
# 这里示例将0-1类合并为0，2-3类合并为1，4-5类合并为2
y_train_3class = y_train_idx // 2
y_train_3class = tf.keras.utils.to_categorical(y_train_3class, num_classes=3)

# 测试集做同样转换
y_test_idx = np.argmax(y_test, axis=1)
y_test_3class = y_test_idx // 2
y_test_3class = tf.keras.utils.to_categorical(y_test_3class, num_classes=3)

# 使用新标签拟合模型
his = model.fit(X_train, y_train_3class, epochs=30, batch_size=32, 
          validation_data=(X_test, y_test_3class), callbacks=[logger])

内容的提问来源于stack exchange，提问作者abcd1211231