[Python] Numpy manual

from

ndarray

basic operation

1
2
3
4
5
6
7
8
9
10
11
12

from numpy import *
a = arange(15).reshape(3, 5)
b = array([[1,2,3], [2,3,4]])
a.shape		# (3, 5)
a.size
a.max(axis=?)
a.mean(axis=?)
# use ravel to replace flatten
a.ravel()
a.T			# transpose
a.astype(_)		# _:	np.float, np.int, np.uint8
a.dtype.name

creation

arange, random.rand, random.randint, linspace, ones, zeros, empty, logspace

slice

• a[…, i]

  … represent unspecified dim

• np.split(a, N)

  split a into evenly N parts

• np.split(a, [])

  split a at certain index

concat

• np.vstack(a, b)
• np.hstack(a, b)
• np.concatenate(a, b, axis = ?)
• np.stack(a, b) Note: Stack in a new dimension
• np.fliplr()
• np.flipud()

basic math operation

• np.dot

  Note: calculate dot product of two vectors

• power

1
2
3
4

from numpy import *
a = ndarray
b = ndarray 	# same size as a
a ** b

random

1
2
3
4
5
6
7
8

import python.random as random

random.seed(CERTAIN_NUMBER)
random.random(SHAPE)
random.randint(LOWER, HIGHER, NUMBER)

b = random.permutation(a)
random.shuffle(a)		# in-place

useful

• operation on mask ndarray

1
2
3
4
5

a = np.arange(10)
a_g = a > 6
a_s = a < 7
a_g + a_s	# or operand
a_g * a_s	# and operand

Numpy Cheatsheet

Those two are copied from BHu by me and the origin of which are shown in the pics.

• ← Previous Post
• Next Post →