Chapter 2: NumPy Quiz

NumPy Quiz written in the style of a kind but demanding teacher who really wants you to understand — not just memorize — NumPy.

I’ve divided it into 5 difficulty levels, from warm-up to expert.

Each question has:

• clear instructions
• expected output format (when applicable)
• hints (only look after trying!)
• full solution + explanation after each block

Instructions for you (the student):

• Try to solve every question without looking at the solution first
• Write your code in a notebook or editor
• Compare your result with the solution
• If you get stuck → read the hint
• If still stuck → look at the solution, but then immediately re-do the question yourself without copying

Ready? Let’s begin.

Level 1 – Warm-up (Basic creation & properties)

Q1. Create four different NumPy arrays that all contain exactly the numbers 0 through 9 (inclusive). Use four different creation methods.

Q2. Create a 6×8 array filled with the number 99. Then change its data type to int16. Finally print: shape, ndim, size, itemsize, nbytes, dtype.

Q3.

Without typing any numbers manually, print:

• the shape
• the number of bytes one element occupies
• the total memory used by the array in bytes
• the data type

Level 2 – Indexing & Slicing (very important)

Q4.

Extract and show (print) the following sub-arrays:

a) the bottom-right 4×4 corner b) every second row and every second column (checkerboard style – starting from 0,0) c) the anti-diagonal (from top-right to bottom-left) d) all elements that are multiples of 11

Q5. Given the same arr:

Replace every number that is divisible by 7 with −999.

Then replace every number that is greater than 90 with 999.

Do both replacements in one single code block (you may use multiple lines).

Q6. Create a 10×10 array filled with zeros. Then set the border (first and last row + first and last column) to 1. Do it without loops and without hard-coding indices

Level 3 – Broadcasting & Arithmetic

Q7. Create a 12×12 multiplication table (1×1 to 12×12) using broadcasting only. Do not use np.meshgrid, np.outer, or any loops.

Q8.

Normalize each column so that it has:

• mean = 0
• standard deviation = 1

Do it in one line.

Then normalize each row the same way (in one line).

Q9. Create a 100×100 array where:

• element [i,j] = sin(i) + cos(j)
• i and j are 0-based indices

Do it without loops and without meshgrid.

Level 4 – Reductions, Masks, Where, Unique

Q10.

Create a new array containing only the rows where:

• the row mean is between 40 and 60 (inclusive)
• AND at least 3 values are positive

Q11. Given the same data:

Replace all values outside the range [0, 100] with the nearest boundary (0 or 100).

Do it in one expression using np.clip or np.where.

Q12.

Compute the mean and standard deviation of values for each label 0–4.

Do it without loops and without pandas.

Q13.

Find the 10 most frequent values and their counts.

Show them sorted by frequency (descending).

Level 5 – Expert / Interview-style questions

Q14. Create a 10×10 array where the value at position [i,j] is the GCD of (i+1) and (j+1).

Q15. Given a 1D array of length 1_000_000 filled with random integers 1–100:

a) Find how many numbers appear exactly 10 000 times b) Find the 10 numbers that appear most frequently (with their counts)

Do both without loops and without pandas.

Q16. Create a function that takes a 2D array and returns a boolean mask of the same shape where:

• True = the element is a local maximum (greater than all its 4 direct neighbors — up, down, left, right)
• False = otherwise

Handle borders correctly (border cells have fewer neighbors).

Do not use loops.

Q17. Given a 1D array a of length N:

Create a 2D array of shape (N−4, 5) where each row is a sliding window of 5 consecutive elements from a.

Do it without loops (use only views / strides if possible).

Q18. (very hard) Given a 2D array of shape (m, n):

Replace every element with the median of its 3×3 neighborhood (including itself). Handle borders by padding with the nearest value (replicate padding).

Do not use loops or scipy.signal.

Now — try to solve at least 5–6 questions before looking at solutions.

When you’re ready, just say:

• “Show solutions for questions 1–10”
• “I want solutions only for the hard ones (14–18)”
• “Give me hints for 16–18 first”
• “Create even harder questions”
• or any other request

I’m here — let me know how you want to proceed! 😊