25 interactive multiple-choice questions with instant feedback, explanations, and lesson links for topics you miss.
~55 min read + exercises
Module 1 quiz and review
Before we begin
Research on learning shows that retrieving an answer (even guessing first) helps memory stick better than re-reading alone. This lesson uses multiple choice questions for that reason.
How to get the most from it:
Read the question fully before looking at options.
Pick an answer — commit before clicking.
Read the feedback whether you were right or wrong (wrong answers teach more when you review the linked lesson).
Aim for at least 19 out of 25 before the project — but honesty beats a high score on the first try.
You can click Try again to reset and retake the whole quiz.
When you finish, you will see a score summary and links to lessons for any topics you missed.
Multiple choice quiz
Interactive quiz
Pick one answer per question. Feedback appears immediately — take your time before clicking.
0 / 25 correct·0 answered
Question 1 of 25
What does a single grayscale pixel store in the computer?
Question 2 of 25
You flatten an 8×6 grayscale patch row by row. How many numbers are in the final list?
Question 3 of 25
Which statement best describes a vector in this course?
Question 4 of 25
What is the dot product of [1, 2, 3] and [4, 5, 6]?
Question 5 of 25
Two flattened patches produce a large positive dot product. What does that usually suggest?
Question 6 of 25
Why is cosine similarity often better than a raw dot product for comparing image patches?
Question 7 of 25
You photograph the same white wall twice. Pixel values differ slightly (142 vs 139). Why?
Question 8 of 25
Why does training use average error over many pixels instead of one pixel only?
Question 9 of 25
On an error graph, the slope at your current weight is negative. What should you do to reduce error?
Question 10 of 25
Training error goes 2.0 → 0.1 → 5.0 → 20 in four steps. What is the most likely cause?
Question 11 of 25
You flatten a 16×16 RGB patch (one list per pixel channel). How long is the list?
Question 12 of 25
In gradient descent, why is there a minus sign in: new weight = old weight − learning rate × slope?
Question 13 of 25
What is a model in machine learning?
Question 14 of 25
What happens during training?
Question 15 of 25
You use ChatGPT to ask a question and get a reply. Which phase is that?
Question 16 of 25
What is loss (error) during training?
Question 17 of 25
A model scores 99% on training data but 50% on new photos. What is the most likely problem?
Question 18 of 25
What is one epoch?
Question 19 of 25
A 32×32 RGB image (3 channels) stored as a matrix has shape:
Question 20 of 25
Learning rate is too small. Training loss barely moves over many steps. What is likely wrong?
Question 21 of 25
Inference means:
Question 22 of 25
In supervised learning, a label is:
Question 23 of 25
On the loss curve, the slope at your current weight is positive. To reduce loss you should:
Question 24 of 25
Which best describes a matrix vs a vector?
Question 25 of 25
Why do we minimize loss during training?
After the quiz
Scored 19/25 or higher? You are in good shape for the project.
Scored lower? That is normal on a first pass. Use the “Topics to review” links in your results, re-read those lessons, then click Try again.
Quick mental checklist before coding:
I can explain model, training, and inference in my own words.
I know what loss, labels, and epochs mean.
I can explain vector vs matrix with an image example.
I can compute a small dot product by hand.
I know why we average error over many pixels.
I can describe gradient descent in a few sentences.
I know what happens when learning rate is too high.