让我们看看分类标签数据:
print (y_train.shape) # (60000,) print (y_train[:10]) # [5 0 4 1 9 2 1 3 1 4]
可以看到,这是一个一位数组,包含了训练数据集对应的标签。
在Keras中,模型训练时,需要把分类标签数据转换为类似位图的矩阵,例如y_train前面10个值是:
[5 0 4 1 9 2 1 3 1 4]
表示为矩阵(对应位置标为1)
[[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.] # 5 对应位置标为1 [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.] # 0 [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.] # 4 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] # 1 [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.] # 9 以此类推 [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.] [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.] [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]]
可以使用to_categorical
来转换标签数据。
# 将一维类数组转换为10维分类矩阵 Y_train = np_utils.to_categorical(y_train, 10) Y_test = np_utils.to_categorical(y_test, 10)
打印Y_train矩阵的形状
print (Y_train.shape) # (60000, 10)
打印Y_train矩阵的前面10条内容:
>>> print (Y_train[:10]) [[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.] [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.] [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.] [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.] [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.] [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]]
可以看到对应 [5 0 4 1 9 2 1 3 1 4]
。