Content: - The exam covers material through the "05 Basic Practice, Part 2" lecture. - Material from lecture notes, lecture, code demos, quizzes, and homework could all show up. Rules: - The exam is a paper exam given during class. - It is closed book but you may bring one 8.5x11" sheet of notes with writing or printing on both sides. - It requires a scientific calculator that can do arithmetic including finding a square root and a base-2 logarithm. - It is a 75-minute exam given during class. I will hand out exams before class, start the exam at the start of class, and end the exam at the end of class. I am strict about collecting exams on time and will penalize students who turn exams in late. (The alternatives are unfair to students who turn exams in on time and to students who have to leave immediately to get across campus for another class or exam.) (For students who have a McBurney letter recommending 1.5 time or other testing accommodations, I have already emailed instructions soon on using the Testing Center.) - You must work on the exam by yourself. - You may not use a laptop or phone or other communication device. - You must not discuss the exam with other students, or with anybody else, via any communication medium until after the exam day. ---------------------------------------------------------------------- Grading and regrade requests: - The exam has 100 points spread across many question parts of a few points each. Some will be graded right or wrong with no partial credit. Some will be graded with partial credit. - If you want to challenge our grading, you must let your TA know in writing within one week of us returning graded exams. - Here is an example of a challenge we will reject: "I did everything right until I skipped a minus sign on a question that was graded right or wrong. Can I have partial credit?" - Here is an example of a challenge we might approve: "I think you made a mistake in grading #4a ... ." (We do not plan to increase partial credit in response to debate, but we want to correct mistakes we made in applying our rubric.) ---------------------------------------------------------------------- Students commonly and reasonably ask "What is on the exam?" Here are some thoughts on this question: - Please see the "Content" block, above. - Several recent exams are posted. This term's exam will be different, in that I will write the best exam I can in the time I have, without feeling constrained by past exams. - There will be some "compute by hand" questions. I include these because my understanding of an algorithm usually goes way up when I run it by hand. - There will be some conceptual questions. e.g. - What effect does increasing ||w|| have on the width of the "road" in hard-margin SVM? - Here is a graph of some red and blue dots ... which value of C will provide the widest road in soft-margin SVM? Which value of C will provide the best training accuracy? - Here are the steps in a demonstration that the negative log likelihood in logistic regression is ... (the formula), and here are some reasons in scrambled order ...; match the reasons to justify the steps. - Here is a graph of some data and a model ... match the Python code to its output, where code options include model.predict(X), model.predict_proba(X), model.score(X, y) and output options include .... - Here is a paragraph describing some data and a scientific question. Which of the five supervised learning algorithm(s) we've seen would be appropriate? (There must be matching questions related to this too.)