Numpy Basics¶

• Fixed Type
• Contiguous Memory

In [ ]:

import numpy as np

import numpy as np

The Basics¶

In [ ]:

a = np.array([1, 2, 3], dtype=np.int8)
b = np.array([[9, 8, 7, 11, 24], [6, 5, 4, 45, 67], [3, 2, 1, 16, 12]], dtype=np.float64)
a, b

a = np.array([1, 2, 3], dtype=np.int8) b = np.array([[9, 8, 7, 11, 24], [6, 5, 4, 45, 67], [3, 2, 1, 16, 12]], dtype=np.float64) a, b

Out[ ]:

(array([1, 2, 3], dtype=int8),
 array([[ 9.,  8.,  7., 11., 24.],
        [ 6.,  5.,  4., 45., 67.],
        [ 3.,  2.,  1., 16., 12.]]))

In [ ]:

# Get dimensions
a.ndim, b.ndim

# Get dimensions a.ndim, b.ndim

Out[ ]:

(1, 2)

In [ ]:

# Get shape
a.shape, b.shape

# Get shape a.shape, b.shape

Out[ ]:

((3,), (3, 5))

In [ ]:

# Get type
a.dtype, b.dtype

# Get type a.dtype, b.dtype

Out[ ]:

(dtype('int8'), dtype('float64'))

In [ ]:

# Each item size
a.itemsize, b.itemsize # each item size in Byte

# Each item size a.itemsize, b.itemsize # each item size in Byte

Out[ ]:

(1, 8)

In [ ]:

# Total size
# nbytes = size * itemsize
a.nbytes, b.nbytes

# Total size # nbytes = size * itemsize a.nbytes, b.nbytes

Out[ ]:

(3, 120)

In [ ]:

# b[i, j] i=row, j=col
# Get specific row
b[1, :]

# b[i, j] i=row, j=col # Get specific row b[1, :]

Out[ ]:

array([ 6.,  5.,  4., 45., 67.])

In [ ]:

# Get even indexes started from index 2
b[:, 2::2] = 78

# Get even indexes started from index 2 b[:, 2::2] = 78

In [ ]:

# b[0, :] = 78
b[0, :] = [1, 2, 3, 4, 5] # List should be the same size and dimensions
b

# b[0, :] = 78 b[0, :] = [1, 2, 3, 4, 5] # List should be the same size and dimensions b

Out[ ]:

array([[ 1.,  2.,  3.,  4.,  5.],
       [ 6.,  5., 78., 45., 78.],
       [ 3.,  2., 78., 16., 78.]])

3D example¶

In [ ]:

c = np.zeros(shape=(2, 3, 4), dtype=np.float16)
c[0, 1, :] = [1, 2, 3, 4]
c

c = np.zeros(shape=(2, 3, 4), dtype=np.float16) c[0, 1, :] = [1, 2, 3, 4] c

Out[ ]:

array([[[0., 0., 0., 0.],
        [1., 2., 3., 4.],
        [0., 0., 0., 0.]],

       [[0., 0., 0., 0.],
        [0., 0., 0., 0.],
        [0., 0., 0., 0.]]], dtype=float16)

In [ ]:

# outside in
c[0, 1, 3]

# outside in c[0, 1, 3]

Out[ ]:

np.float16(4.0)

Initialize different types of arrays¶

In [ ]:

np.zeros((2, 3))

np.zeros((2, 3))

Out[ ]:

array([[0., 0., 0.],
       [0., 0., 0.]])

In [ ]:

np.ones((2, 3), dtype=np.int8)

np.ones((2, 3), dtype=np.int8)

Out[ ]:

array([[1, 1, 1],
       [1, 1, 1]], dtype=int8)

In [ ]:

np.full(shape=(2, 3), fill_value=123)

np.full(shape=(2, 3), fill_value=123)

Out[ ]:

array([[123, 123, 123],
       [123, 123, 123]])

In [ ]:

np.full_like(a, 12)

np.full_like(a, 12)

Out[ ]:

array([12, 12, 12], dtype=int8)

Random¶

In [ ]:

# Random int
np.random.randint(low=0, high=12, size=(3, 4))

# Random int np.random.randint(low=0, high=12, size=(3, 4))

Out[ ]:

array([[ 2, 10,  3, 11],
       [ 0,  8,  7, 11],
       [ 8,  6,  7,  2]])

In [ ]:

# Random 0-1 just pass size to it
np.random.rand(3, 2)

# Random 0-1 just pass size to it np.random.rand(3, 2)

Out[ ]:

array([[0.95567647, 0.6325864 ],
       [0.7441816 , 0.50506503],
       [0.41507009, 0.04014501]])

In [ ]:

# Random 0-1 just pass size to it
np.random.random_sample(size=b.shape)

# Random 0-1 just pass size to it np.random.random_sample(size=b.shape)

Out[ ]:

array([[0.65321004, 0.93401116, 0.239411  , 0.1699394 , 0.15857918],
       [0.49180305, 0.47913441, 0.67379083, 0.0516425 , 0.68931425],
       [0.78729527, 0.07978258, 0.71997382, 0.55340929, 0.1960921 ]])

In [ ]:

# Identity matrix
np.identity(6)

# Identity matrix np.identity(6)

Out[ ]:

array([[1., 0., 0., 0., 0., 0.],
       [0., 1., 0., 0., 0., 0.],
       [0., 0., 1., 0., 0., 0.],
       [0., 0., 0., 1., 0., 0.],
       [0., 0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 0., 1.]])

In [ ]:

array = np.array([2, 4, 6])
np.repeat(array, 3)

array = np.array([2, 4, 6]) np.repeat(array, 3)

Out[ ]:

array([2, 2, 2, 4, 4, 4, 6, 6, 6])

In [ ]:

array = np.array([[2, 4, 6]])
np.repeat(array, 3, axis=0)

array = np.array([[2, 4, 6]]) np.repeat(array, 3, axis=0)

Out[ ]:

array([[2, 4, 6],
       [2, 4, 6],
       [2, 4, 6]])

Exercise¶

In [ ]:

# Create a zeros array that edges=1 and center=9
# Way 1
array = np.zeros(shape=(11, 11), dtype=np.int8)
r_l, c_l = array.shape

array[r_l // 2, c_l // 2] = 9

array[0::r_l - 1, :] = 1
array[:, 0::c_l - 1] = 1

array

# Create a zeros array that edges=1 and center=9 # Way 1 array = np.zeros(shape=(11, 11), dtype=np.int8) r_l, c_l = array.shape array[r_l // 2, c_l // 2] = 9 array[0::r_l - 1, :] = 1 array[:, 0::c_l - 1] = 1 array

Out[ ]:

array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 9, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], dtype=int8)

In [ ]:

# Way 2
array = np.ones(shape=(11, 11), dtype=np.int8)
r_l, c_l = array.shape

z = np.zeros(shape=(r_l - 2, c_l -2))
z[(r_l - 2) // 2, (c_l - 2) // 2] = 9

array[1:-1, 1:-1] = z

array

# Way 2 array = np.ones(shape=(11, 11), dtype=np.int8) r_l, c_l = array.shape z = np.zeros(shape=(r_l - 2, c_l -2)) z[(r_l - 2) // 2, (c_l - 2) // 2] = 9 array[1:-1, 1:-1] = z array

Out[ ]:

array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 9, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], dtype=int8)

Mathematics¶

In [ ]:

a = np.array([1, 2, 3, 4])
a + 2 # * ** + - / // etc

a = np.array([1, 2, 3, 4]) a + 2 # * ** + - / // etc

Out[ ]:

array([1, 2, 3, 4])

In [ ]:

b = np.array([2, 4, 9, 8]),
a + b # * ** +,,   - / // etc

b = np.array([2, 4, 9, 8]), a + b # * ** +,, - / // etc

Out[ ]:

array([[ 3,  6, 12, 12]])

In [ ]:

np.sin(b)

np.sin(b)

Out[ ]:

array([[ 0.90929743, -0.7568025 ,  0.41211849,  0.98935825]])

Linear Algebra¶

In [ ]:

mat_a = np.random.randint(1, 5, (3, 4))
mat_b = np.random.randint(1, 5, (4, 2))

mat_a, mat_b

mat_a = np.random.randint(1, 5, (3, 4)) mat_b = np.random.randint(1, 5, (4, 2)) mat_a, mat_b

Out[ ]:

(array([[2, 2, 2, 2],
        [4, 4, 2, 1],
        [4, 3, 4, 2]]),
 array([[2, 1],
        [4, 1],
        [3, 3],
        [4, 2]]))

In [ ]:

# mat_a@mat_b
np.matmul(mat_a, mat_b)

# mat_a@mat_b np.matmul(mat_a, mat_b)

Out[ ]:

array([[26, 14],
       [34, 16],
       [40, 23]])

In [ ]:

# Find the determinant
c = np.identity(3)
np.linalg.det(c)

# Find the determinant c = np.identity(3) np.linalg.det(c)

Out[ ]:

np.float64(1.0)

Statistics¶

In [ ]:

stats = np.array([[1,2,3], [4, 5, 6]])

stats = np.array([[1,2,3], [4, 5, 6]])

In [ ]:

np.min(stats, axis=0) # each cols

np.min(stats, axis=0) # each cols

Out[ ]:

array([1, 2, 3])

In [ ]:

np.max(stats, axis=1) # each rows

np.max(stats, axis=1) # each rows

Out[ ]:

array([3, 6])

In [ ]:

np.mean(stats)

np.mean(stats)

Out[ ]:

np.float64(3.5)

In [ ]:

np.sum(stats)

np.sum(stats)

Out[ ]:

np.int64(21)

Reorganizing Arrays¶

In [ ]:

stats.reshape((-1,))

stats.reshape((-1,))

Out[ ]:

array([1, 2, 3, 4, 5, 6])

Vertical & Horizontal stacks¶

In [ ]:

a = stats.reshape((3, 2))
b = np.random.randint(1, 5, (3, 2))

a = stats.reshape((3, 2)) b = np.random.randint(1, 5, (3, 2))

In [ ]:

np.hstack([a, b])

np.hstack([a, b])

Out[ ]:

array([[1, 2, 4, 4],
       [3, 4, 3, 3],
       [5, 6, 3, 3]])

In [ ]:

np.vstack([a, b])

np.vstack([a, b])

Out[ ]:

array([[1, 2],
       [3, 4],
       [5, 6],
       [4, 4],
       [3, 3],
       [3, 3]])

Load Data¶

In [ ]:

# np.genfromtxt("file.txt", delimiter=",")

# np.genfromtxt("file.txt", delimiter=",")

Boolean Masking and Advanced Indexing¶

In [ ]:

data = np.random.randint(1, 500, (5, 4))

data = np.random.randint(1, 500, (5, 4))

In [ ]:

data[data > 100] # return values that match with condition

data[data > 100] # return values that match with condition

Out[ ]:

array([469, 396, 484, 155, 310, 265, 416, 337, 240, 135, 353, 482, 201])

In [ ]:

# data > 100
((data > 100) & (data < 200))
# ~((data > 100) & (data < 200)) # not

# data > 100 ((data > 100) & (data < 200)) # ~((data > 100) & (data < 200)) # not

Out[ ]:

array([[ True,  True,  True,  True],
       [ True, False,  True,  True],
       [ True,  True,  True,  True],
       [ True,  True, False,  True],
       [ True,  True,  True,  True]])

In [ ]:

np.any(data > 469, axis=0) # col

np.any(data > 469, axis=0) # col

Out[ ]:

array([False,  True, False,  True])

Index with list¶

In [ ]:

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

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

Out[ ]:

array([2, 3, 9])

In [ ]: