See https://piazza.com/class/lzsgiv2awue3l6/post/327 for information on taking the exam, including the URL and information about authentication.
There are two examinations planned for this class.
The examinations will be completed remotely. In a previous COVID semester, one of the most common student concerns for remote education was the integrity of assignments: 61% of students agreed with the statement "I am worried about others cheating and my grade looking worse by comparison." As a result, we are taking aggressive, experimental measures with exam creation and delivery to help ensure that student grades reflect mastery of the material.
A distinct examination will be created for each student. This will involve not only usual techniques, such as selecting a subset of questions or answers from a larger pool, but also software-specific approaches. For example, for problems on the examination involving the analysis or manipulation of a program (e.g., "Create a test suite for this Python program that maximizes branch coverage ..."), a different random program will be constructed for each student by the examination server (which will also compute a per-student answer key).
One goal of this process is to reduce the cost/benefit tradeoff for cheating: if completing two exams requires twice as much work and the risk is increased (e.g., the course staff have twice as many datapoints on which to spot suspicious patterns), the incentive to cheat is lessened.
The examinations will be open book, open notes, read-only internet. In essence, you can use any resource except another live human. You can view Stack Overflow, but you can't make new posts. You can use the free version of ChatGPT, but not any paid version of generative AI. You can use MiMaizey, the U-M AI Tool.
The dates for the in-class examinations are posted well in advance. To request a rescheduled examination (e.g., for reasons other than a documented one as above, etc.), you must notify us via email at EECS481Staff@gmail.com (privately) at least one week before the date of the examination. Note that requests need not be granted.
If you have an approved time accommodation form from SSD, complete this 481 SSD Exam Time Confirmation Form as well to confirm that you receive extra time on 481 exams. This "extra step" process helps ensure that we don't forget you.
On the day of the exam, you will be able to choose any two-hour window you like to take the exam. (That is, you can pick any 2 hours within the 24-hour day.)
You can ask clarification questions via Private Piazza Posts on the course forum. On the day of the exam, the forum will be set to Private Posting only.
The interface auto-saves your work every five minutes and you can also save and submit as often as you like. We will grade your last submission before the two-hour deadline.
Because these exams are normally completable in 80 minutes and we are giving 120 minute slots (with 5 extra minutes for technology or network difficulties) within a 24 hour window, we will be very strict about time constraints. Late submissions will not be accepted without significant documentation of an accepted excuse. Please plan ahead accordingly.
Each exam generally covers everything up to and including the day of the exam, except any lecture or reading on the day of the exam itself.
For Fall 2024, the first Exam covers all topics up to Thursday October 3rd except Dataflow Analysis.
The exam does cover the general notion of Static Analysis, and it does cover any concepts related to Static (or Dynamic) Analyses (like false positives and false negatives), but it does not cover the particular technique of Dataflow Analysis.
(Dataflow Analysis will be the first topic on Exam #2.)
Students sometimes wonder if certain lectures, like guest lectures or end-of-term lectures, are included: yes, all types of lectures fair game for the exam. (The "Trivia" and "Psychology" slides with colored borders and the Optional Reading are not part of the exam, but may be part of extra credit questions.) This means that something in the lecture the "day before" the exam is fair game. This also means that the second exam is cumulative and may include material also covered in the first part of the exam (but it is likely to focus more on new material).
General philosophy:
Note that in previous semesters the material was covered in a different order, the exams may have been given in person or via another take-home method, and many other things may have changed. It is often very tempting for students to read the previous exams and then worry: "Oh my, this previous exam is very long and has strange rule X and covers topic Y that I've never heard of. Will my exam be like that?" No, it will not. The previous exams are provided as a courtesy and as a source of example problems: they are not strict guidelines.
When asking questions on the forum about previous exams, we strongly recommend that you include a screenshot of the question, rather than just saying "Q3 of S20". Including a screenshot helps makes sure that the answer will have enough context to help other students as well.
Previous Midterm Exams (i.e., Exam #1) and answer keys are provided as is — many are known to be missing a few problems or to have text cut off in places or to have typos. Unfortunately, we do not have any additional exams (or more complete copies) — everything I have is already posted here.
Previous Final Exams (i.e., Exam #2) and answer keys:
Note that provided answer keys need not be exhaustive: they often list "one possible answer" rather than "all possible answers".
If there is sufficient interest, the TAs will organize review sessions for the exam. This will not happen automatically. Demonstrate your interest via the forum or your discussion section.
"Producing items by means as simple as saying, writing, or typing them can yield substantial memory improvements relative to silent reading. We review the research on this production effect and outline some important extensions and boundary conditions. We also evaluate the evidence that production enhances the distinctiveness of items in memory during encoding, thereby facilitating their later retrieval. There are issues to resolve and areas to explore, but production offers a practical means of enhancing some forms of long-term, explicit memory."
Colin M. MacLeod and Glen E. Bodner. The Production Effect in Memory. Memory, Vol 26, Issue 4, 2017.