NumPy, short for Numerical Python, is a foundational library for scientific computing in Python. Many data analysis packages, including pandas, are built on top of NumPy.

At its core, NumPy uses the ndarray (N-dimensional array) data structure. While similar to Python lists, ndarrays are more efficient due to their C-based implementation and require uniform data types for optimal performance in large-scale mathematical operations.

Installation and Import

Install NumPy using pip:

pip install numpy

Import the library with the conventional alias:

import numpy as np

Creating Arrays

Manual Array Creation

Use np.array() to create arrays from sequences:

# One-dimensional array
vector = np.array([4, 5, 6])

# Two-dimensional array
matrix = np.array([[1, 2, 3], [7, 8, 9]])

Specialized Array Creation Methods

Zero Arrays:

# 1D zero array
zero_vec = np.zeros(5)
print(zero_vec)  # Output: [0. 0. 0. 0. 0.]

# 2D zero array
zero_mat = np.zeros((2, 3))
print(zero_mat)
# Output:
# [[0. 0. 0.]
#  [0. 0. 0.]]

One Arrays:

# 1D array of ones
ones_vec = np.ones(4)
print(ones_vec)  # Output: [1. 1. 1. 1.]

# 2D array of ones
ones_mat = np.ones((3, 2))
print(ones_mat)
# Output:
# [[1. 1.]
#  [1. 1.]
#  [1. 1.]]

Empty Arrays:

# Uninitialized 1D array
empty_arr = np.empty(3)

# Uninitialized 2D array
empty_mat = np.empty((2, 2))

Range Arrays:

# Array from 0 to 9
seq_arr = np.arange(10)
print(seq_arr)  # Output: [0 1 2 3 4 5 6 7 8 9]

# Array from 5 to 15 with step 3
step_arr = np.arange(5, 16, 3)
print(step_arr)  # Output: [ 5  8 11 14]

Random Arrays:

# 2x3 array with random values
rand_mat = np.random.randn(2, 3)
print(rand_mat)
# Output example:
# [[-0.234  1.456 -0.789]
#  [ 0.123 -0.456  0.890]]

Accessing Array Elements

Numeric Indexing

Access elements using zero-based indexing:

arr = np.array([10, 20, 30, 40])
print(arr[2])  # Output: 30

mat = np.array([[1, 2, 3], [4, 5, 6]])
print(mat[1][2])  # Output: 6

Slicing

Extract subarrays with slice notation:

arr = np.array([0, 1, 2, 3, 4, 5])
print(arr[2:5])  # Output: [2 3 4]

mat = np.array([[10, 20, 30], [40, 50, 60]])
print(mat[0:2, 1:3])
# Output:
# [[20 30]
#  [50 60]]

Boolean Indexing

Filter arrays using boolean conditions:

values = np.array([5, 15, 25, 35])
mask = np.array([True, False, True, False])
print(values[mask])  # Output: [ 5 25]

Vectoriaztion and Broadcasting

Vectorization allows element-wise operations without explicit loops:

arr = np.array([[1, 2], [3, 4]])
print(arr + 10)
# Output:
# [[11 12]
#  [13 14]]

arr1 = np.array([1, 2, 3])
arr2 = np.array([4, 5, 6])
print(arr1 * arr2)  # Output: [ 4 10 18]

Broadcasting automatically aligns arrays of different shapes:

vec = np.array([1, 2, 3])
mat = np.array([[10, 20, 30], [40, 50, 60]])
print(vec + mat)
# Output:
# [[11 22 33]
#  [41 52 63]]

Common Array Methods and Properties

Array Attributes

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

print(arr.ndim)    # Dimensions: 2
print(arr.shape)   # Shape: (2, 3)
print(arr.size)    # Total elements: 6
print(arr.dtype)   # Data type: int32

Statistical Operations

arr = np.array([10, 20, 30, 40])

print(arr.max())    # Maximum: 40
print(arr.min())    # Minimum: 10
print(arr.mean())   # Mean: 25.0
print(arr.sum())    # Sum: 100

Sorting

NumPy provides in-place and non-destructive sorting options:

arr = np.array([30, 10, 40, 20])

# NumPy sort (returns new array)
sorted_np = np.sort(arr)
print(sorted_np)  # Output: [10 20 30 40]

# Python built-in sort
sorted_py = sorted(arr, reverse=True)
print(sorted_py)  # Output: [40, 30, 20, 10]

Conditional Filtering

Combine boolean indexing with comparison operators:

arr = np.array([5, 15, 25, 35])

# Single condition
print(arr[arr > 20])  # Output: [25 35]

# Multiple conditions
print(arr[(arr > 10) & (arr < 30)])  # Output: [15 25]

Transposition

Swap array axes using the .T attribute:

mat = np.array([[1, 2, 3], [4, 5, 6]])
print(mat.T)
# Output:
# [[1 4]
#  [2 5]
#  [3 6]]