Introduction à NumPy

NumPy Installation and Introduction

• NumPy is installed using pip install numpy
• NumPy provides a powerful array object, ndarray, more efficient than Python lists, especially for large datasets.

Importing NumPy

• Import NumPy using import numpy as np

Basic Arrays (ndarrays)

• ndarray is a N-dimensional, homogenous array (all elements are the same data type).

Creating NumPy Arrays

• From a Python list:

  • 1D array: array_1d = np.array([1, 2, 3, 4])
  • 2D array: array_2d = np.array([[1, 2, 3], [4, 5, 6]])

• Using NumPy Functions:

  • Array of zeros: zeros_array = np.zeros((3, 4)) (3x4 array filled with zeros)
  • Array of ones: ones_array = np.ones((2, 3)) (2x3 array filled with ones)
  • Array with a constant value: full_array = np.full((2, 2), 7) (2x2 array filled with 7)

Array Properties

• Shape: array_2d.shape (e.g., (2, 3))
• Size: array_2d.size (e.g., 6)
• Data type: array_1d.dtype (e.g., dtype('int64'))
• Number of dimensions: array_2d.ndim (e.g., 2)

Indexing and Slicing

• Similar to Python lists, but with more capabilities for multidimensional arrays.
• Simple indexing: element = array_2d[0, 1] (access element at row 0, column 1)
• Slicing: sub_array = array_2d[0:2, :2] (first 2 rows, first 2 columns)
• Boolean Masking: mask = array_1d > 2 creates a boolean mask; filtered_array = array_1d[mask] filters elements greater than 2.

Basic Operations

• Element-wise operations:

  • Addition: result = array + 2
  • Multiplication: result = array * 2
  • Exponentiation: result = array ** 2

• Operations between arrays (element-wise):

  • Addition: result = a + b

• Universal Functions (ufuncs): performs element-wise calculations (e.g., np.sqrt, np.sum, np.mean, np.prod)

Array Manipulation

• Reshaping: reshaped = array_2d.reshape((3, 2)) (changes the shape of the array)
• Concatenation:
  • Horizontal: concat_array = np.hstack((a, b))
  • Vertical: concat_array = np.vstack((a, b))
• Transposition: transposed = array_2d.T

Linear Algebra

• Matrix Product: product = np.dot(A, B)

Error Handling and Missing Values

• Managing NaN (Not a Number) and inf (infinity) values in NumPy arrays.
• Ignoring NaN values during calculations: mean_without_nan = np.nanmean(nan_array)

Differences with Python Lists

• Speed: NumPy operations are significantly faster than Python list operations, especially for large datasets.
• Memory Efficiency: NumPy arrays use less memory than Python lists for large datasets.
• Vectorized Operations: NumPy operations are vectorized, working on entire arrays at once, eliminating the need for explicit loops.