NumPy Q&A 20 Core Questions
Interview Prep

NumPy for ML & Scientific Computing: Interview Q&A

Short questions and answers on NumPy: ndarrays, broadcasting, vectorization and basic linear algebra.

ndarrays Broadcasting Vectorization Linear Algebra
1 What is a NumPy ndarray and why is it useful? âš¡ Beginner
Answer: An ndarray is a multi-dimensional, homogeneous array providing fast vectorized operations and serving as the base for many ML libraries.
2 How do you create a NumPy array from a Python list? âš¡ Beginner
Answer: Use np.array(my_list).
3 What are shape and dtype attributes of an array? âš¡ Beginner
Answer: shape gives the dimensions (rows, cols, ...), dtype gives the data type of elements.
4 What is broadcasting in NumPy? 📊 Intermediate
Answer: Broadcasting lets arrays of different but compatible shapes be combined in arithmetic by implicitly expanding dimensions.
5 Why is vectorization preferred over Python loops in NumPy code? 📊 Intermediate
Answer: Vectorized operations run in optimized C code, making them much faster and more concise than explicit Python loops.
6 How do you compute basic statistics like mean and std in NumPy? âš¡ Beginner
Answer: Use np.mean(a), np.std(a), optionally with axis to aggregate along dimensions.
7 How do you perform matrix multiplication with NumPy arrays? âš¡ Beginner
Answer: Use the @ operator or np.dot(A, B) / A.dot(B).
8 What is the difference between view() and copy() in NumPy? 🔥 Advanced
Answer: A view shares the same underlying data; a copy allocates new memory. Changing a view affects the original array.
9 How do you reshape an array? âš¡ Beginner
Answer: Use a.reshape(new_shape) or a.reshape(-1, n) to infer one dimension.
10 What is axis in aggregation functions like sum or mean? 📊 Intermediate
Answer: axis specifies which dimension to collapse; e.g., axis=0 aggregates rows, axis=1 aggregates columns in 2D arrays.
11 How can you generate random numbers with NumPy? âš¡ Beginner
Answer: Use np.random module or the np.random.default_rng() generator with methods like normal, integers, choice.
12 What is boolean indexing in NumPy? 📊 Intermediate
Answer: It selects elements where a boolean mask is True, e.g., a[a > 0].
13 How do you stack arrays vertically and horizontally? 📊 Intermediate
Answer: Use np.vstack([a, b]) for vertical, np.hstack([a, b]) for horizontal, or np.concatenate with axis.
14 How can you compute eigenvalues and eigenvectors with NumPy? 🔥 Advanced
Answer: Use np.linalg.eig(A) for general matrices or np.linalg.eigh for symmetric/Hermitian ones.
15 What is the difference between np.dot and elementwise * multiplication? 📊 Intermediate
Answer: * multiplies elements position-wise; dot (or @) performs matrix/inner product.
16 Why is NumPy so important for machine learning in Python? âš¡ Beginner
Answer: Because libraries like pandas, scikit-learn, PyTorch, TensorFlow all rely on efficient NumPy-style arrays and operations.
17 What are some common performance pitfalls when using NumPy? 🔥 Advanced
Answer: Pitfalls: using Python loops instead of vectorization, excessive copies, not using views, mixing Python lists and arrays.
18 How do you set the random seed for reproducible NumPy results? âš¡ Beginner
Answer: With the legacy API: np.random.seed(42); with Generator API: rng = np.random.default_rng(42).
19 Give a real-world example where NumPy is heavily used in an ML project. âš¡ Beginner
Answer: NumPy underpins feature preprocessing, custom loss functions, numerical simulations and glue code between ML libraries.
20 What is the key message to remember about NumPy for ML engineers? âš¡ Beginner
Answer: NumPy is the numerical backbone of the Python ML stack; fluent array thinking and broadcasting will make your ML code faster and cleaner.

Quick Recap: NumPy

Mastering ndarrays, indexing, broadcasting and basic linear algebra gives you the core numerical skills behind most Python ML frameworks.

Back: Pandas Q&A Next: Mixed ML Q&A - 1