mean() and median() in PyTorch

By /

mean() can get the mean(average) values as shown below:

*Memos:

  • mean() can be called both from torch and a tensor.
  • The 2nd argument is one or more dimensions with torch.
  • The 1st argument is one or more dimensions with a tensor.
  • mean() can only accept floating-point or complex numbers so you need conversion to them if they are not as I explain it in my answer(17.1) otherwise there is the error.
import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.mean(my_tensor.float())tensor(5.4167)
my_tensor.float().mean()
# tensor(5.4167)

torch.mean(my_tensor.float(), 0)
my_tensor.float().mean(0)
torch.mean(my_tensor.float(), (0,))
my_tensor.float().mean((0,))
torch.mean(my_tensor.float(), -2)
my_tensor.float().mean(-2)
torch.mean(my_tensor.float(), (-2,))
my_tensor.float().mean((-2,))
# tensor([4.6667, 5.6667, 6.3333, 5.0000])

torch.mean(my_tensor.float(), 1)
my_tensor.float().mean(1)
torch.mean(my_tensor.float(), (1,))
my_tensor.float().mean((1,))
torch.mean(my_tensor.float(), -1)
my_tensor.float().mean(-1)
torch.mean(my_tensor.float(), (-1,))
my_tensor.float().mean((-1,))
# tensor([5.7500, 4.7500, 5.7500])

torch.mean(my_tensor.float(), (0, 1))
my_tensor.float().mean((0, 1))
torch.mean(my_tensor.float(), (0, -1))
my_tensor.float().mean((0, -1))
torch.mean(my_tensor.float(), (1, 0))
my_tensor.float().mean((1, 0))
torch.mean(my_tensor.float(), (1, -2))
my_tensor.float().mean((1, -2))
torch.mean(my_tensor.float(), (-1, 0))
my_tensor.float().mean((-1, 0))
torch.mean(my_tensor.float(), (-1, -2))
my_tensor.float().mean((-1, -2))
torch.mean(my_tensor.float(), (-2, 1))
my_tensor.float().mean((-2, 1))
torch.mean(my_tensor.float(), (-2, -1))
my_tensor.float().mean((-2, -1))
# tensor(5.4167)
Enter fullscreen mode
Exit fullscreen mode
import torch my_tensor = torch.tensor([[[0., 1., 2.], [3., 4., 5.]], [[6., 7., 8.], [9., 10., 11.]], [[12., 13., 14.], [15., 16., 17.]], [[18., 19., 20.], [21., 22., 23.]]])
torch.mean(my_tensor)
my_tensor.mean()
# tensor(11.5000)

torch.mean(my_tensor, 0)
my_tensor.mean(0)
torch.mean(my_tensor, (0,))
my_tensor.mean((0,))
# tensor([[9., 10., 11.], [12., 13., 14.]])

torch.mean(my_tensor, 1)
my_tensor.mean(1)
torch.mean(my_tensor, (1,))
my_tensor.mean((1,))
# tensor([[1.5000, 2.5000, 3.5000], [7.5000, 8.5000, 9.5000],
# [13.5000, 14.5000, 15.5000], [19.5000, 20.5000, 21.5000]])

torch.mean(my_tensor, 2)
torch.mean(my_tensor, (2,))
my_tensor.mean(2)
my_tensor.mean((2,))
torch.mean(my_tensor, -1)
my_tensor.mean(-1)
torch.mean(my_tensor, (-1,))
my_tensor.mean((-1,))
# tensor([[1., 4.], [7., 10.], [13., 16.], [19., 22.]])

torch.mean(my_tensor, -2)
my_tensor.mean(-2)
torch.mean(my_tensor, (-2,))
my_tensor.mean((-2,))
# tensor([[1.5000, 2.5000, 3.5000],
# [7.5000, 8.5000, 9.5000],
# [13.5000, 14.5000, 15.5000],
# [19.5000, 20.5000, 21.5000]])

torch.mean(my_tensor, -3)
my_tensor.mean(-3)
torch.mean(my_tensor, (-3,))
my_tensor.mean((-3,))
# tensor([[ 9., 10., 11.], [12., 13., 14.]])

torch.mean(my_tensor, (0, 1))
my_tensor.mean((0, 1))
torch.mean(my_tensor, (1, 0))
my_tensor.mean((1, 0))
torch.mean(my_tensor, (1, -3))
my_tensor.mean((1, -3))
torch.mean(my_tensor, (0, -2))
my_tensor.mean((0, -2))
torch.mean(my_tensor, (-2, 0))
my_tensor.mean((-2, 0))
torch.mean(my_tensor, (-2, -3))
my_tensor.mean((-2, -3))
torch.mean(my_tensor, (-3, 1))
my_tensor.mean((-3, 1))
torch.mean(my_tensor, (-3, -2))
my_tensor.mean((-3, -2))
# tensor([10.5000, 11.5000, 12.5000])

torch.mean(my_tensor, (0, 2))
my_tensor.mean((0, 2))
torch.mean(my_tensor, (0, -1))
my_tensor.mean((0, -1))
torch.mean(my_tensor, (2, 0))
my_tensor.mean((2, 0))
torch.mean(my_tensor, (2, -3))
my_tensor.mean((2, -3))
torch.mean(my_tensor, (-1, 0))
my_tensor.mean((-1, 0))
torch.mean(my_tensor, (-1, -3))
my_tensor.mean((-1, -3))
torch.mean(my_tensor, (-3, 2))
my_tensor.mean((-3, 2))
torch.mean(my_tensor, (-3, -1))
my_tensor.mean((-3, -1))
# tensor([10., 13.]) 
torch.mean(my_tensor, (1, 2))
my_tensor.mean((1, 2))
torch.mean(my_tensor, (1, -1))
my_tensor.mean((1, -1))
torch.mean(my_tensor, (2, 1))
my_tensor.mean((2, 1))
torch.mean(my_tensor, (2, -2))
my_tensor.mean((2, -2))
torch.mean(my_tensor, (-1, 1))
my_tensor.mean((-1, 1))
torch.mean(my_tensor, (-1, -2))
my_tensor.mean((-1, -2))
torch.mean(my_tensor, (-2, 2))
my_tensor.mean((-2, 2))
torch.mean(my_tensor, (-2, -1))
my_tensor.mean((-2, -1))
# tensor([2.5000, 8.5000, 14.5000, 20.5000])

torch.mean(my_tensor, (0, 1, 2))
my_tensor.mean((0, 1, 2))
etc.
# tensor(11.5000)
Enter fullscreen mode
Exit fullscreen mode

median() can get the median(average) values as shown below:

*Memos:

  • median() can be called both from torch and a tensor.
  • The 2nd argument is one or more dimensions with torch.
  • The 1st argument is one or more dimensions with a tensor.
import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.median(my_tensor)
my_tensor.median()
# tensor(5)

torch.median(my_tensor, 0)
my_tensor.median(0)
torch.median(my_tensor, -2)
my_tensor.median(-2)
# torch.return_types.median(
# values=tensor([5, 5, 7, 5]),
# indices=tensor([0, 1, 0, 1]))

torch.median(my_tensor, 1)
my_tensor.median(1)
torch.median(my_tensor, -1)
my_tensor.median(-1)
# torch.return_types.median(
# values=tensor([5, 5, 3]),
# indices=tensor([0, 1, 3]))
Enter fullscreen mode
Exit fullscreen mode
import torch my_tensor = torch.tensor([[[0., 1., 2.], [3., 4., 5.]], [[6., 7., 8.], [9., 10., 11.]], [[12., 13., 14.], [15., 16., 17.]], [[18., 19., 20.], [21., 22., 23.]]])
torch.median(my_tensor)
my_tensor.median()
# tensor(11.)

torch.median(my_tensor, 0)
my_tensor.median(0)
# torch.return_types.median(
# values=tensor([[6., 7., 8.], [9., 10., 11.]]),
# indices=tensor([[1, 1, 1], [1, 1, 1]]))

torch.median(my_tensor, 1)
my_tensor.median(1)
# torch.return_types.median(
# values=tensor([[0., 1., 2.], [6., 7., 8.],
# [12., 13., 14.], [18., 19., 20.]]),
# indices=tensor([[0, 0, 0], [0, 0, 0],
# [0, 0, 0], [0, 0, 0]]))

torch.median(my_tensor, 2)
my_tensor.median(2)
torch.median(my_tensor, -1)
my_tensor.median(-1)
# torch.return_types.median(
# values=tensor([[1., 4.], [7., 10.], [13., 16.], [19., 22.]]),
# indices=tensor([[1, 1], [1, 1], [1, 1], [1, 1]]))

torch.median(my_tensor, -2)
my_tensor.median(-2)
# torch.return_types.median(
# values=tensor([[0., 1., 2.], [6., 7., 8.],
# [12., 13., 14.], [18., 19., 20.]]),
# indices=tensor([[0, 0, 0], [0, 0, 0],
# [0, 0, 0], [0, 0, 0]]))

torch.median(my_tensor, -3)
my_tensor.median(-3)
# torch.return_types.median(
# values=tensor([[6., 7., 8.], [9., 10., 11.]]),
# indices=tensor([[1, 1, 1], [1, 1, 1]]))
Enter fullscreen mode
Exit fullscreen mode

Discover more from Coursity

Subscribe to get the latest posts sent to your email.

Leave a Comment

Your email address will not be published. Required fields are marked *

Discover more from Coursity

Subscribe now to keep reading and get access to the full archive.

Continue reading

Scroll to Top
Making an svg image from scratch : a step by step guide web develop mentor. Boost your wordpress website traffic with the martinstools ad network plugin. Implementing these strategies, you can ensure that your membership website thrives in 2023.