Please leave me comment below if you have any questions, if you notice any errors or if you would like to discuss a topic further.

Questions and Comments

This slecture is reviewed by Anantha Raghuraman

This slecture is very interesting. Soonam first discusses about false alarm rate and miss rate, which are terms generally used by engineers and talks about how these are similar to the type I and type II errors in statistics. He then follows it up to discuss about mis-classification error and how the mis-classification error compares with the false-alarm rate/miss rate and type I and type II errors. This serves as a very nice introduction to the Neyman Pearson Lemma that he would go on to discuss.

He then goes on to talk about the mathematical formulation of the Neyman-Pearson Lemma and then talks about how all of what he had talked about can be used to draw ROC curves. Finally, he talks about the the key factors that one should look at, while drawing ROC curves. That is, what constitutes a good ROC curve and what does not.

Overall, the slecture is very compact and contains a lot of insightful material. It was a pleasure to read this as a follow up to what was taught in the ECE 662 course.

Suggestions:

a) I would suggest that the author gives a little more background information to the topics that he discusses (for example, he could consider explaining what a randomized likelihood ratio test (LRT) is, very briefly). The sole purpose is, I feel that it would be much better if this slecture is somewhat more self-contained, than how it is now.

b) To motivate the problem better, since this comes under the pattern recognition course, the author could have spent a little more time discussing about why and where we use the concepts discussed here.

c) Finally, a very minor suggestion is that, the author could consider formatting the lecture uniformly. (For example, make the pictures, of similar dimension and size), just for better presentation.

Overall, it is full of useful content, but requires you to have enough statistics background or willingness to read background material.

Back to Neyman-Pearson Lemma and Receiver Operating Characteristic Curve