Python学习之旅——numpy库基础使用总结（超详细）

介绍

基础用法

矩阵创建

numpy矩阵的类型为numpy.ndarray，
没有指定数据类型，默认是float64类型

基础创建

a = np.array([[1,2,3], [4,5,6], [7,8,9], [10, 11, 12]])
b = np.arange(4)

列表与numpy矩阵的转换：

x = [675.99524, 166.36523, 691.63257, 193.72832]
y = np.asarray(x)

print(x, type(x))
print(y, type(y), y.shape)

numpy与list的转换

np.arange(largest_recall, 1, 0.01).tolist()

特殊矩阵

全零矩阵：

## 创建(2,)维的全0向量
d = np.zeros(2)
print(d, d.shape)

d = np.zeros((2,1))
print(d, d.shape)

d = np.zeros((3,4))
print(d, d.shape)

# 下面两个构造方法等价
d = np.zeros((10, 3, 3), dtype=np.uint8)
d = np.zeros([10, 3, 3], dtype=np.uint8)

注意不同矩阵的维度区别。

全1矩阵：

b = np.ones((1,2))

常数矩阵：

c = np.full((2,2), 7)  # Create a constant array
print(c)               # Prints "[[ 7.  7.]
                       #          [ 7.  7.]]"

单位矩阵：

d = np.eye(2)         # Create a 2x2 identity matrix
print(d)              # Prints "[[ 1.  0.]
                      #          [ 0.  1.]]"

随机矩阵：

# 创建指定维度的随机矩阵
x = np.random.rand(4,3)
x = np.random.random([4,3])
x = np.random.random([4,3,2])

y = np.random.randint(0,10,(4,3)) # [0,10)区间, shape为(4,3)的随机矩阵

x1 = np.random.uniform(-1,1) # 指定区间均匀分布随机数

矩阵元素数据类型修改

没有指定数值类型时，numpy矩阵有默认数值类型：

x = np.array([1, 2, 3]) # dtype('int64')
x = np.array([1.0, 2, 3]) dtype('float64')

修改数值类型方法：

x = x.astype(float)
x = x.astype(bool)
x = np.array(x, dtype=float)

矩阵切片

一个矩阵的切片是对相同矩阵元素数据的不同观察角度，共享相同的数据，修改切片后矩阵会影响原矩阵。

a = np.array([[1,2,3], [4,5,6], [7,8,9], [10, 11, 12]])

print(a)  # prints "array([[ 1,  2,  3],
          #                [ 4,  5,  6],
          #                [ 7,  8,  9],
          #                [10, 11, 12]])"

# Create an array of indices
b = np.array([0, 2, 0, 1])

print(a[np.arange(4), b])  # Prints "[ 1  6  7 11]"

索引切片会降维，但是列表索引或范围索引不会降维：

overlap_0_7 = np.array([[0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7],
                            [0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7],
                            [0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7]])
overlap_0_5 = np.array([[0.7, 0.5, 0.5, 0.7, 0.5, 0.5, 0.5, 0.5],
                            [0.5, 0.25, 0.25, 0.5, 0.25, 0.5, 0.5, 0.5],
                            [0.5, 0.25, 0.25, 0.5, 0.25, 0.5, 0.5, 0.5]])
min_overlaps = np.stack([overlap_0_7, overlap_0_5], axis=0)  # [2, 3, 8]
print(min_overlaps, min_overlaps.shape)
print(len(min_overlaps))

current_classes = [0]
min_overlaps = min_overlaps[:, :, current_classes] # [2, 3, 1]
print(min_overlaps, min_overlaps.shape)

矩阵运算

矩阵解包

d = np.zeros(2)
print(d, d.shape)

x, y = d
print("x, y: ", x, y)

常用内置函数

np.asarray函数

np.ascontiguousarray函数

np.copy函数

np.hstack和np.vstack函数

np.logical_and函数

np.min和np.max函数

求解指定维度上最小或最大值，返回值直接就是得到的最小最大值，会进行降维。

x = np.random.randint(0, 10, (3, 8))
print(x, x.shape)
y = np.max(x, 1)
print(y)

结果：

[[3 0 3 9 6 9 9 6]
 [8 0 3 1 4 7 6 8]
 [5 7 2 8 4 0 5 8]] (3, 8)
[9 8 8]

np.reshape函数

https://stackoverflow.com/questions/39549331/reshape-numpy-n-vector-to-n-1-vector?rq=1

np.stack函数

overlap_0_7 = np.array([[0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7],
                            [0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7],
                            [0.7, 0.5, 0.5, 0.7, 0.5, 0.7, 0.7, 0.7]])
overlap_0_5 = np.array([[0.7, 0.5, 0.5, 0.7, 0.5, 0.5, 0.5, 0.5],
                            [0.5, 0.25, 0.25, 0.5, 0.25, 0.5, 0.5, 0.5],
                            [0.5, 0.25, 0.25, 0.5, 0.25, 0.5, 0.5, 0.5]])
min_overlaps = np.stack([overlap_0_7, overlap_0_5], axis=0)  # [2, 3, 8]