Quiz: SVMs & Perceptron

Methods of AI — SoSe 2026


Q1 — SVM

Question: What are support vectors, and why are they the only points that matter for defining the hyperplane?

Max’s answer:
Result:


Q2 — SVM

Question: What is the kernel trick? Why is it useful?

Max’s answer:
Result:


Q4 — SVM

Question: What is the margin in SVM? Write the formula and explain what it means geometrically.

Max’s answer:
Result:


Q5 — SVM

Question: What is the perceptron, and why can’t it solve XOR?

Max’s answer:
Result:


Q6 — SVM

Question: Name three common kernel functions and give their formulas.

Max’s answer:
Result:


Q7 — SVM

Question: Write the SVM optimization problem (what we minimize and what the constraint is).

Max’s answer:
Result:


Beyond the lecture (optional)

These questions go beyond the SoSe 2026 lecture slides (textbook / external additions). Kept for depth, not exam-critical.

Q3 — SVM

Question: What does the C parameter in soft-margin SVM control? What happens with large vs. small C?

Max’s answer:
Result:


Session 25/05/26 — slide-level rapid recall ⚡

5 questions on the core slide-level SVM mechanics (margin, support vectors, kernel trick, decision rule, convexity). Hard difficulty.

S1 — Margin

Question: The SVM margin is 2/‖w‖. Where does the 2 come from — why not just 1/‖w‖?

Max’s answer:
Result:


S2 — Support vectors

Question: You delete a training point and re-train. When does the hyperplane stay identical, and when does it move? Tie your answer to a formula.

Max’s answer:
Result:


S3 — Kernel trick

Question: State precisely what K(x, x′) computes and why we never have to build the feature map Φ.

Max’s answer:
Result:


S4 — Decision rule

Question: Write the kernelized decision function for a new input x, and explain how the class is read off it.

Max’s answer:
Result:


S5 — Convexity

Question: Why does the SVM optimization have a single, unique solution (unlike training a perceptron or neural net)?

Max’s answer:
Result:


Score

Total: / 7
Session 25/05/26 (slide-level): / 5