Design and Analysis of Algorithms (4 cu)
Period I, Fall 2012

Summary and comments on course feedback
November 1, 2012 (MK), revised December 11, 2012 (MK)


Revision comments:

By Dec 11, no new entries were received via the course feedback system. 

Ten students attended the renewal exam held on Nov 27; one of them passed 
the course. The other nine students got notably low number of points. This 
was rather surprising, given that two exam problems were identical to two 
regular exercises given in the course book (15.4-5 and 35.3-2). It seems 
like many attendees of the exam had not really practiced solving exercises, 
or even read the descriptions of the exercises. This, again, suggests that 
the instructor needs to stress that one may not expect to pass the course
without reading the material and practicing with the exercises.


Basic information: 

The course belongs to advanced studies and is mandatory to students in 
the specialization area of algorithms and machine learning. The course 
takes a traditional form: six weeks of lectures, 4 hours per week; 
weekly exercise sessions; a 2.5-hour exam (held on Oct 18). The course 
follows very closely selected chapters of the course book (CLRS, 3rd 
ed.). Lectures are given mostly using a black board, essentially no 
lecture slides are made available.

A large number of students (50) registered for the course, 35 attended 
at least some of the course activities. The course exam was attended 
by 26 students, of which 15 passed the course, with grade distribution 
(5*** 4*** 3**** 2** 1***). The renewal exam will take place on Nov 27 
---at the moment, 16 students have registered for the renewal exam. 


Student feedback:

By Nov 1, we received a total of 10 entries via the anonymous course 
feedback system. The variance among the entries is generally very high 
for most of the six feedback questions. Regarding the clarity of 
learning objectives and the supportiveness of learning material, the 
happiness was nearly uniformly distributed over the entire range from 
1 (very unhappy) to 5 (very happy). Regarding the supportiveness of 
the activities and the assessment of the course, the mode was at the 
happier side (concentrating heavily at 4). However, in almost all 
feedback entries the course was marked as quite laborious (mode at 4-5). 
As a whole, the course was graded as follows: 5*, 4***, 3***, 2, 1***, 
the bimodality showing fairly strong polarity among the respondents. 

While the verbal comments also show high variability, reflecting the 
numeric scores, a couple of issues occur repeatedly. First, many wished 
that some lecture notes or slides were available. Second, some 
respondents found the lectures, to some extent, hand-wavy, expressing 
preference for more rigorous treatment. In addition, a few complained 
about the exam that it was not possibly to solve all the problems in 
the given time, unless you see almost immediately how to solve them.


Reflections by the instructor:

The feedback seems to stem from the students' heterogeneous background 
and somewhat unjustified expectations. Many of the respondents seem to 
have not taken seriously the central role of the course book: almost 
all the exercises were taken from the book; many of the exam questions 
were taken from the book; lecture slides or notes are not made available 
simply because they would just repeat what is in the book. The material 
covered in the course occupy about 250-280 pages of the book 
(including exercises and problems, excluding what is assumed as 
prerequisite). This should be a fair number of pages, for the exposition 
in the book is not very dense but quite gentle.

There also seems to be some misunderstanding concerning the nature of 
lectures. Namely, there are good reasons why much of the technicalities 
are skipped in a hand-wavy manner during the lectures, but not in the 
exercises (or in the exam): (a) It is assumed that the student already 
masters the needed math routines (e.g., induction proofs)---they must 
have been learned in basic math courses and prior CS courses (in the 
data structures course in particular). (b) Some are slower and some are 
faster at technical reasoning. Therefore, everybody should take his/her 
individual time, which comes ideally true in solving exercise problems. 

Finally, the exam questions are really not puzzles but variants of 
exercise problems or examples from the book, testing how well the 
student is able to recognize the design or analysis method in question. 
Note that six small problems are not heavier than four larger problems. 
However, it seems that a typical student has somewhat undeveloped 
skills for managing a situation where the task is to solve as much as 
possible in a limited time. The exam questions were designed so that 
one should be able to solve 50% of every problem right away, so that 
the majority of the time remains for completing answers to some, if 
not all, the problems. Note that an exam is not very good if every 
other student is able to solve all the problems (nearly) correctly in 
the given time. From the tester's point of view, having many and 
difficult questions means playing in the safer side.


Suggested changes for the next realization of the course:

1. Stress the importance and role of the course book. "Most probably 
you will not pass the course without it".

2. Go through the main themes and learning objectives very explicitly. 
Announcing the content of the course and having a link to the themes 
and objectives matrix seem not sufficient.

3. Announce the covered parts of the course book still in more detail. 
Make explicit what pages are assumed as prior knowledge, and what 
(additional) pages belong to the course requirements (material for the 
exam). This may mean listing some single pages of the book.

4. Make an old exam and its model solutions available at least a week 
before the course exam. Go through a couple of exam questions in the 
last lecture week. In fact, it might be a good idea to show old exam 
questions already in the very beginning of the course, and then come 
back to them at the end of the course.

5. In the exam, consider having one question less - even then the 
resolution is expected to remain high enough concerning higher grades.

6. To lighten the workload of the course, consider leaving out one 
exercise problem per week (i.e., have 5 instead of 6), and perhaps 
going through those left-overs in the last 2x45-min lecture session.