Performance of ML Models
15 Points
Explain in your own words the meanings of the following terms in the context of performance of Machine Learning models. Please note that you are to write what you understand these terms measure or express rather than repeating the definition in books or on the internet.
(i) Confusion Matrix
(ii) Accuracy
(iii) Precision
(iv) Recall
(v) F1-Score
3/6
ROC and AUC
Read up on ROC and AUC concepts in Machine learning from any book or the Internet.
(i) Explain in your own words, what is your interpretation of the utility of ROC curve? (Again please understand what it is rather than responding with a verbatim copy from the book/internet).
(ii) What does AUC measure? If the AUC of an ML model is 1.0 – what would that mean about the ML model?
4/6
Fundamentals of Logistic Regression
(i) Let’s assume that the probability of a binary under investigation being malware is 0.8. What are the odds in favor of it being a malware?
(ii) The logit function takes as input probability p of a random item being in a particular class, and outputs the log of the odds of that item being in that class. Suppose that the item is characterized by n features expressed by a feature vector x1, x2, …,xn . Assuming that log of odds for the item to be in the special class is a linear function of its features, we can say that logit(p) = w1x1 + w2x2 + … + wn*xn + b.
If w1x1 + + w2x2 + … + wn*xn + b is given, compute p as a function of wi’s, b, and xi’s
5/6
Regularization
(i) Read up on the concepts of bias and variance in the context of ML models. Explain your own interpretation of these terms.
(ii) Explain the idea of regularization in ML models — what does it mean to you intuitively?
Q5 SVM
5 Points
In the context of ML, what is meant by margin? Why do we want to maximize the margin during classification?
6/6