Probability & Statistics Quiz

Probability & Statistics Quiz

Work through these questions to check your understanding of the probabilistic and statistical foundations of machine learning.

Exercise

X is 0 with probability 0.5 and 10 with probability 0.5. What is E[X]?

Need a hint?

Probability-weighted average.

Exercise

A Gaussian with σ = 0.1 has PDF value ≈ 3.99 at its peak. Is this a valid probability density?

Need a hint?

What must equal 1 — the values, or the area?

Exercise

A disease affects 2% of people. A test catches 90% of cases but false-positives on 10% of healthy people. You test positive — roughly what is the chance you're sick?

Need a hint?

Weigh true positives (0.9 × 0.02) against false positives (0.1 × 0.98).

Exercise

You assume Laplace-distributed noise on your regression targets and fit by maximum likelihood. Which loss are you effectively minimizing?

Need a hint?

The Laplace log-density contains |y − ŷ|.

Exercise

How does the MAP objective differ from the MLE objective?

Need a hint?

Take the log of posterior ∝ likelihood × prior.

Exercise

Which distribution over 4 outcomes has the MAXIMUM entropy?

Need a hint?

Entropy measures unpredictability.

Exercise

Your estimate of a mean has a standard error you'd like to cut in half. By what factor must you increase the sample size?

Need a hint?

SE = σ/√n, so SE depends on √n.

Exercise

An A/B test on 5 million users finds a 0.01% conversion lift with p = 0.001. What is the right conclusion?

Need a hint?

Statistical significance and practical significance are different things.

Exercise

In the ML project loop, deriving a loss function as the negative log-likelihood of the data (MLE) belongs to which slot?

Need a hint?

MLE tells you what 'good' means — the quantity to minimize.

Exercise

Transfer test — a technique this course did NOT teach: walk-forward (time-based) cross-validation instead of random k-fold. Which slot of the ML project loop does it change?

Need a hint?

It changes how you estimate generalization, not the model or the loss.