ValueError：层顺序的输入 0 与层不兼容：：预期 min_ndim=3，发现 ndim=2。收到的完整形状：(10, 24)

Question

import numpy as np
import pandas as pd
from numpy.random import seed
import tensorflow as tf
from tensorflow import keras
from keras import Sequential
from keras.layers import Dense, Conv1D, MaxPooling2D, Activation
from sklearn.model_selection import train_test_split

seed(1)
tf.random.set_seed(2)
droprate = 0.5

dataset = pd.read_csv('filecounts.csv')
data = np.array(pd.get_dummies(dataset['counts']))

model = Sequential()

model.add(Conv1D(8, kernel_size=3, padding="same", activation="relu",input_shape=(12, 12, 10)))
model.add(MaxPooling2D(pool_size=2))
...
model.add(Conv1D(4, kernel_size=3, padding="same", activation="relu"))
model.add(MaxPooling2D(pool_size=2))
...
model.add(Conv1D(1, kernel_size=3, padding="same", activation="relu"))
model.add(MaxPooling2D(pool_size=2))
...
model.add(Activation("softmax"))

sgd = keras.optimizers.SGD(learning_rate=1)

train, test = train_test_split(data, test_size=0.5)

model.compile(optimizer=sgd, loss='binary_crossentropy', metrics=['accuracy'])
model.fit(train, epochs=100, batch_size=10)
_, accuracy = model.evaluate(test, verbose=0, steps=1)

print('Accuracy: %.2f' % (accuracy*100))

Answer 1

Conv1D 期望输入形状为 3+D 张量，形状为：batch_shape + (steps, input_dim)，输出形状为：batch_shape + (new_steps, filters) 的 3+D 张量，带或不带 padding='same'。

错误是由于 Maxpooling2D 需要形状为 (batch_size, rows, cols, channels) 的 4D 张量。

工作示例代码

import tensorflow as tf
import numpy as np
import tensorflow.keras as keras

X_train = np.random.random((12,12,10))
y_train = np.random.random((12, 1))

model = tf.keras.Sequential()

model.add(keras.layers.Conv1D(8, kernel_size=3, padding="same", activation="relu",input_shape=(12, 10)))
model.add(keras.layers.MaxPool1D(pool_size=2))

model.add(keras.layers.Conv1D(4, kernel_size=3, padding="same", activation="relu"))
model.add(keras.layers.MaxPool1D(pool_size=2))

model.add(keras.layers.Conv1D(1, kernel_size=3, padding="same", activation="relu"))
model.add(keras.layers.MaxPool1D(pool_size=2))
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(units = 128, activation = 'relu'))
model.add(keras.layers.Dense(units = 1, activation = 'softmax'))

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

model.fit(X_train, y_train, epochs = 15)

输出

Epoch 1/10
1/1 [==============================] - 1s 944ms/step - loss: 0.6914 - accuracy: 0.0000e+00
Epoch 2/10
1/1 [==============================] - 0s 9ms/step - loss: 0.6900 - accuracy: 0.0000e+00
Epoch 3/10
1/1 [==============================] - 0s 9ms/step - loss: 0.6885 - accuracy: 0.0000e+00
Epoch 4/10
1/1 [==============================] - 0s 7ms/step - loss: 0.6870 - accuracy: 0.0000e+00
Epoch 5/10
1/1 [==============================] - 0s 8ms/step - loss: 0.6856 - accuracy: 0.0000e+00
Epoch 6/10
1/1 [==============================] - 0s 9ms/step - loss: 0.6841 - accuracy: 0.0000e+00
Epoch 7/10
1/1 [==============================] - 0s 8ms/step - loss: 0.6828 - accuracy: 0.0000e+00
Epoch 8/10
1/1 [==============================] - 0s 14ms/step - loss: 0.6814 - accuracy: 0.0000e+00
Epoch 9/10
1/1 [==============================] - 0s 7ms/step - loss: 0.6801 - accuracy: 0.0000e+00
Epoch 10/10
1/1 [==============================] - 0s 11ms/step - loss: 0.6789 - accuracy: 0.0000e+00
<keras.callbacks.History at 0x7eff56169810>