- Unsupervised learning is the most common alternative when supervised learning is not applicable. Is it correct?
- The CEO of your company asks you to find out the factors that determined a negative sales trend. What kind of analysis do you need to perform?
- Given a dataset of independent samples and a candidate data generating process (for example, a Gaussian distribution), the likelihood is obtained by summing the probabilities of all samples. It is correct?
- Under which hypothesis can the likelihood be computed as a product of single probabilities?
- Suppose we have a dataset of students containing some unknown numerical features (for example, age, marks, and so on). You want to separate male and female students, so you decide to cluster the dataset into two groups. Unfortunately, both clusters have roughly 50% male and 50% female students. How can you explain this result?
- Consider the previous example, but repeat the experiment and cluster into five groups. What do you expect to find in each of them? (List some reasonable possibilities.)
- You've clustered the customers of an online store. Given a new sample, what kind of prediction can you make?
