keras to_categorical 增加了额外的价值

Question

我有 4 个类需要预测，我正在使用 keras 的 to_categorical 来实现它，我希望得到一个 4 one-hot-encoded 数组，但似乎我得到了5 个值，而是为所有行

显示一个额外的 [0] 值

dict = {'word': 1, 'feature_name': 2, 'feature_value': 3, 'part_number': 4}
    Y = dataset['class'].apply(lambda label: dict[label])
    print(Y.unique()) #prints [1 4 2 3]
    train_x, test_x, train_y, test_y = model_selection.train_test_split(X, Y, test_size=0.2, random_state=0)
    train_y = to_categorical(train_y)
    print(train_y[0])# prints [0. 0. 1. 0. 0.]

正在尝试构建的模型如下

model = Sequential()
model.add(Dense(10, input_dim=input_dim, activation='relu'))
model.add(Dense(10, activation='relu'))
model.add(Dense(10, activation='relu'))
model.add(Dense(10, activation='relu'))
model.add(Dense(4, activation='softmax'))

但它一直在抛出

ValueError: Error when checking target: expected dense_5 to have shape (4,) but got array with shape (5,)

Answer 1

可能是keras版本错误。尝试更新它，因为这对我有用：

dict = {'word': 1, 'feature_name': 2, 'feature_value': 3, 'part_number': 4}
Y = np.random.randint(4, size=10)
print(np.unique(Y)) #prints [0 1 2 3]
train_y = np_utils.to_categorical(Y, num_classes=4)
print(train_y[0]) # prints [0. 0. 1. 0.]

尝试从 0 开始你的字典，因为当 Keras 读取你的数据时以 0 作为参考。

dict = {'word': 0, 'feature_name': 1, 'feature_value': 2, 'part_number': 3}

如果不行，试试强制类的数字：

train_y = to_categorical(train_y, num_classes = 4)

Answer 2

您需要从 0 开始编号类，像这样：

dict = {'word': 0, 'feature_name': 1, 'feature_value': 2, 'part_number': 3}

您可以通过 help() 命令获取函数的描述

help(np_utils.to_categorical)

:

Help on function to_categorical in module keras.utils.np_utils:

to_categorical(y, num_classes=None, dtype='float32')
Converts a class vector (integers) to binary class matrix.

E.g. for use with categorical_crossentropy.

# Arguments
    y: class vector to be converted into a matrix
        (integers from 0 to num_classes).
    num_classes: total number of classes.
    dtype: The data type expected by the input, as a string
        (`float32`, `float64`, `int32`...)

# Returns
    A binary matrix representation of the input. The classes axis
    is placed last.

keras to_categorical 增加了额外的价值

keras to_categorical adds additional value

keras

one-hot-encoding