TensorFlow - ValueError：形状 (3, 1) 和 (4, 3) 不兼容

Question

我对 DL 完全陌生，当我适合我的模型时，我遇到了这个错误

ValueError: Shapes (3, 1) and (4, 3) are incompatible

数据集：

Features: [0.22222222 0.625      0.06779661 0.04166667], Target: [1 0 0]
Features: [0.16666667 0.41666667 0.06779661 0.04166667], Target: [1 0 0]
Features: [0.11111111 0.5        0.05084746 0.04166667], Target: [1 0 0]
Features: [0.08333333 0.45833333 0.08474576 0.04166667], Target: [1 0 0]
Features: [0.19444444 0.66666667 0.06779661 0.04166667], Target: [1 0 0]

型号：

def build_fc_model():
  fc_model = tf.keras.Sequential([
      tf.keras.layers.Dense(4, activation=tf.nn.softmax),
      tf.keras.layers.Dense(4, activation=tf.nn.softmax),
      tf.keras.layers.Dense(3, activation=tf.nn.softmax),
  ])
  return fc_model```

model.fit 出错

model = build_fc_model()
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=1e-1), loss='categorical_crossentropy', metrics=['accuracy'])

BATCH_SIZE = 10
EPOCHS = 5

model.fit(dataset, batch_size=BATCH_SIZE, epochs=EPOCHS)

感谢您的帮助

Answer 1

在您的代码中，build_fc_model 中缺少 InputLayer，因此请检查一下：

import tensorflow as tf
import numpy as np

def build_fc_model():
  fc_model = tf.keras.Sequential([
      tf.keras.layers.InputLayer((4,)),
      tf.keras.layers.Dense(4, activation=tf.nn.softmax),
      tf.keras.layers.Dense(4, activation=tf.nn.softmax),
      tf.keras.layers.Dense(3, activation=tf.nn.softmax),
  ])
  return fc_model


data = np.array([[0.22222222, 0.625,      0.06779661, 0.04166667], 
                 [0.16666667, 0.41666667, 0.06779661, 0.04166667],
                 [0.11111111, 0.5 ,       0.05084746, 0.04166667], 
                 [0.08333333, 0.45833333, 0.08474576, 0.04166667], 
                 [0.19444444, 0.66666667, 0.06779661, 0.04166667]])

target = np.array([[1, 0 ,0],
                   [1, 0 ,0],
                   [1, 0 ,0],
                   [1, 0 ,0],
                   [1, 0 ,0]])

model = build_fc_model()
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=1e-1), loss='categorical_crossentropy', metrics=['accuracy'])

BATCH_SIZE = 1
EPOCHS = 5

model.fit(data, target, batch_size=BATCH_SIZE, epochs=EPOCHS)

输出：

Epoch 1/5
5/5 [==============================] - 0s 991us/step - loss: 0.8198 - accuracy: 0.6000
Epoch 2/5
5/5 [==============================] - 0s 603us/step - loss: 0.1590 - accuracy: 1.0000
Epoch 3/5
5/5 [==============================] - 0s 593us/step - loss: 0.0372 - accuracy: 1.0000
Epoch 4/5
5/5 [==============================] - 0s 597us/step - loss: 0.0131 - accuracy: 1.0000
Epoch 5/5
5/5 [==============================] - 0s 680us/step - loss: 0.0064 - accuracy: 1.0000