张量的数学计算

求和

二维张量求和：

A=torch.arange(20,dtype=torch.float32).reshape(5,4)
print(f'A.shape: {A.shape}')

A_sum_axis0=A.sum(axis=0)
print(f'A_sum_axis0: {A_sum_axis0}')
print(f'A_sum_axis0.shape: {A_sum_axis0.shape}')

A_sum_axis1=A.sum(axis=1)
print(f'A_sum_axis1: {A_sum_axis1}')
print(f'A_sum_axis1.shape: {A_sum_axis1.shape}')

三维张量求和：

B=torch.arange(20,dtype=torch.float32).reshape(2,2,5)
print(B)
B_sum_axis0=B.sum(axis=0)
print(f'B_sum_axis0: {B_sum_axis0}')
print(f'B_sum_axis0.shape: {B_sum_axis0.shape}')

B_sum_axis1=B.sum(axis=1)
print(f'B_sum_axis1: {B_sum_axis1}')
print(f'B_sum_axis1.shape: {B_sum_axis1.shape}')

B_sum_axis2=B.sum(axis=2)
print(f'B_sum_axis2: {B_sum_axis2}')
print(f'B_sum_axis2.shape: {B_sum_axis2.shape}')

求均值

print(f'A.mean(): {A.mean()}')
print(f'A.sum() / A.numel() : {A.sum() / A.numel()}')

print(f'A.mean(axis=0): {A.mean(axis=0)}')
print(f'A.sum(axis=0) / A.shape[0] : {A.sum(axis=0) / A.shape[0]}')

非降维求和

sum_A=A.sum(axis=0,keepdims=True)
print(f'sum_A: {sum_A}')
print(f'sum_A.shape: {sum_A.shape}')

哈达玛积

对应位置元素相乘称为哈达玛积：

print(f'B: {B}')
print(f'A*B: {A*B}')

点积

相同位置的按元素乘积的和（等价于先按元素相乘，再对结果求和）：

x=torch.arange(4,dtype=torch.float32)
y=torch.ones(4,dtype=torch.float32)
print(x)
print(y)
print(torch.dot(x,y))
print(torch.sum(x*y))

矩阵向量积

print(torch.mv(A,x))

矩阵-矩阵乘法

print(torch.mm(A,B) # 等价于 torch.matmul())