NumPy Arrays

By the end of this lesson, you will be able to create, inspect, and perform basic vectorized operations on NumPy arrays to accelerate numerical data analysis.

What it is

A NumPy array (ndarray) is a grid of values, all of the same type, indexed by a tuple of non-negative integers. Unlike Python lists, which can hold mixed types and store pointers to objects, NumPy arrays store contiguous blocks of raw binary data in memory. This structure allows for highly optimized mathematical operations using C-level loops rather than slow Python interpreters.

Key terms include shape (dimensions), dtype (data type), and vectorization (applying an operation to entire arrays at once).

Why it matters

• Performance: Operations are 10-100x faster than equivalent Python list comprehensions due to memory locality and compiled code.
• Conciseness: Complex mathematical expressions become single-line statements without explicit loops.
• Interoperability: It is the standard data structure for libraries like Pandas, Scikit-Learn, and TensorFlow.
• Memory Efficiency: Storing homogeneous data reduces overhead compared to Python objects.

Syntax or steps

1. Import the library: import numpy as np.
2. Create an array from a list: np.array([1, 2, 3]).
3. Inspect properties: use .shape, .dtype, and .size.
4. Perform element-wise math: arr * 2 multiplies every element by 2.

Example

import numpy as np

# Create a 1D array
scores = np.array([85, 90, 78, 92, 88])

# Basic inspection
print(f"Shape: {scores.shape}")
print(f"Dtype: {scores.dtype}")

# Vectorized operation: Normalize scores to 0-1 range
min_val = scores.min()
max_val = scores.max()
normalized = (scores - min_val) / (max_val - min_val)

print(f"Normalized: {normalized}")

# Boolean indexing: Find scores above average
avg = scores.mean()
above_avg = scores[scores > avg]
print(f"Above Average: {above_avg}")

Explanation:

• np.array(...) converts a Python list into a fixed-type integer array.
• scores.min() and scores.max() compute statistics efficiently.
• The normalization line uses broadcasting; the scalar results of min_val and max_val are applied to every element in scores simultaneously.
• scores > avg creates a boolean mask array. Passing this mask back into scores filters the original array, returning only elements where the condition was true.

Common mistakes

• Mixed Types: If you pass [1, "two", 3] to np.array, NumPy casts everything to strings (<U21 dtype), breaking math operations. Always ensure input data is numeric if intended for calculation.
• In-place Modification Confusion: arr + 1 creates a new array. To modify the original, use arr += 1.
• Indexing Errors: NumPy uses zero-based indexing like Python, but negative indices wrap around. Ensure bounds checks when slicing large datasets.
• Ignoring Dtype: Integer division truncates results. Use astype(float) before dividing if you need decimal precision.

When to use it

Practice

Guided Exercise: Create an array of numbers 1 through 10. Calculate the square of each number using vectorization (no loops). Print the result.

Hint: Use np.arange(1, 11) and multiply the array by itself (arr * arr or arr ** 2).

Challenge: Given an array of temperatures in Celsius, convert them to Fahrenheit using the formula $F = C \times 1.8 + 32$. Filter out any temperatures below freezing (0°C).

Quick check

Q: What happens if you try to add two NumPy arrays with different shapes that cannot be broadcast together?

A: A ValueError is raised because the dimensions are incompatible for element-wise operations.

Summary

NumPy arrays provide a high-performance foundation for numerical computing by storing homogeneous data contiguously and enabling vectorized operations. Mastering array creation, inspection, and broadcasting is essential for efficient data analytics and machine learning workflows.