numpy.mean, numpy.std의 axis 알아보기

첫번째, 다음의 홈페이지를 방문하여 읽어본다.

numpy. xxx 설명 페이지

두번째, 위 페이지를 읽고 작성한 나의 코드는 다음과 같다.

# 평균과 표준편차는 채널별로 구해줍니다.
x_mean = np.mean(x_train, axis=(0, 1, 2))
x_mean_dft = np.mean(x_train)
x_mean_0 = np.mean(x_train, axis=0)
x_mean_1 = np.mean(x_train, axis=1)
x_mean_2 = np.mean(x_train, axis=2)
x_mean_3 = np.mean(x_train, axis=3)
x_mean_01 = np.mean(x_train, axis=(0, 1))
x_mean_02 = np.mean(x_train, axis=(0, 2))
x_mean_03 = np.mean(x_train, axis=(0, 3))
x_mean_012 = np.mean(x_train, axis=(0, 1, 2))
x_mean_013 = np.mean(x_train, axis=(0, 1, 3))
x_std = np.std(x_train, axis=(0, 1, 2))
print(x_mean.ndim)
print("x_train.shape: ", x_train.shape)      # (50000, 32, 32, 3)
print("x_mean_dft.shape: ", x_mean_dft.shape)
print("x_mean_0.shape: ", x_mean_0.shape)   # (32, 32, 3) 첫번째 50000 사라짐
print("x_mean_1.shape: ", x_mean_1.shape)  # (50000, 32, 3) 두번째 32 사라짐
print("x_mean_2.shape: ", x_mean_2.shape)  # (50000, 32, 3) 세번재 32 사라짐
print("x_mean_3.shape: ", x_mean_3.shape)   # (50000, 32, 32) 네번째 3 사라짐
print("x_mean_01.shape: ", x_mean_01.shape)  # 50000, 32 사라짐
print("x_mean_02.shape: ", x_mean_02.shape)  # 50000, 32 사라짐
print("x_mean_03.shape: ", x_mean_03.shape)  # 50000, 3 사라짐
print("x_mean_012.shape: ", x_mean_012.shape) # 50000, 32, 32 없어짐
print("x_mean_013.shape: ", x_mean_013.shape) # 50000, 32, 3 없어짐
print(x_std.shape)
"""

결과

1
x_train.shape:  (50000, 32, 32, 3)
x_mean_dft.shape:  ()
x_mean_0.shape:  (32, 32, 3)
x_mean_1.shape:  (50000, 32, 3)
x_mean_2.shape:  (50000, 32, 3)
x_mean_3.shape:  (50000, 32, 32)
x_mean_01.shape:  (32, 3)
x_mean_02.shape:  (32, 3)
x_mean_03.shape:  (32, 32)
x_mean_012.shape:  (3,)
x_mean_013.shape:  (32,)

실험 결과를 보고, numpy.sum 등의 axis에 대해 고민해보자.