Laboratory of Computer and Information Science / Neural Networks Research Centre

Helsinki University of Technology → Department of Computer Science and Engineering →
Laboratory of Computer and Information Science → Teaching →
T-61.3050 Machine Learning: Basic Principles → 2007 → Mid-course survey

Mid-course survey 2007 - T-61.3050

We asked the students on the 23 October 2007 lecture, at which point the lectures were half way, to fill in the anonymous survey form. 15 returned the filled form after the lecture.

Below you can find a summary of the answers.

The answers have been somewhat normalized (for example, zeros have been converted to ones when the scale has been 1-5). Empty items ("don't know", no reply) have been omitted from the analysis.


How familiar are the following concepts for you? (on a scale of 1 to 5, 1 being "not familiar" and 5 being "familiar")

mean1st qu.median3rd qu.
ComputationalComplexity3.32.544
ConvergenceOfAlgorithms2.1123
LinearDiscrimination1.9123
DecisionTrees2.0123
ClusteringAlgorithms2.1123
StatisticalTests2.0113

Comment: Purpose of this question was to test the prior knowledge on the topics yet to be discussed in the course. Most of the students have some idea of the topics, but there are many for which the topics are completely new, with the expectation of the rather well known computational complexity.


Please grade the following parts of the course on a scale of 1 to 5, 1 being "bad" and 5 being "good"; "-" being "don't know":

mean1st qu.median3rd qu.
LecturesGrade3.1333.5
ProblemSessionsGrade3.433.54
TextBookGrade3.4334
OtherStudyMaterialGrade2.9233
OrganizatioOfTheCourseGrade3.9345
InformationOfTheCourseGrade4.1444.5
UsefulnessOfTheCourseGrade4.03.2544.75
SelectionOfTopicsGrade4.1444.75
OverlapWithOtherCoursesGrade3.4334
ObjectivesOfTheCourseGrade3.93.2544.75
GradingCriteriaGrade3.73.544
CorrespondanceReqsVsGradingCriteriaGrade3.6344
CourseAsAWholeGrade3.6334

Comment: The course got a grade of 3.6. Especially the selection of topics and usefulness of the course was appreciated.


How important are the following in understanding the course? (on a scale of 1 to 5, 1 being "not important" and 5 being "important")

mean1st qu.median3rd qu.
Lectures3.9345
SolvingProblems4.3455
ProblemSessions4.3455
LectureSlides3.7345
TextBook4.0445
CommunicationWithFellowStudents2.8234
TermProject4.1445
OtherNANANANA

Comment: The applied part of the course, that is, the problems and term project, are perceived important in understanding this course. This is no surprise for a course that covers mathematical methods. Relatively, communication with fellow students was not deemed important, even though discussing and working the problems with others is often quite effective. In future courses we might try to encourage some group activity (although probably not require it). It was mentioned that information found elsewhere in the web may also be helpful.


How many credits should a student awarded for passing this course (you'll get 5)?

mean1st qu.median3rd qu.
ShouldGetCredits5.8556

How demanding is the course? (on a scale of 1 to 5, 1 being "too easy", 3 being "about right" and 5 being "too demanding")

mean1st qu.median3rd qu.
HowDemanding3.8344

Comment: The course is perceived to be rather demanding than easy. Nobody thought that it is on the side of too easy...


Please answer the following questions on a scale of 1 to 5, 1 being "no" and 5 being "yes"; "-" being "don't know":

mean1st qu.median3rd qu.
HavePrerequisites4.13.555
PrerequisitesAppropriate4.2455
EnglishDifficult2.0113
FeelThatLearned2.82.2533
KnowWhatImportant2.822.54
GoodAtmosphere4.1445
DifferenceBtwClearIrrelevant2.41.523
DifficultPresentedUnderstandably2.51.252.53
PartsSupport3.333.54

Comment: The students mostly had the prerequisite knowledge (not having prerequisite knowledge actually had some correlation with the perceived demandingness of the course). One thing that we need to work out is to improve the score on learning the topics and identification of important parts. Hopefully, however, this will improve as the course proceeds (there are more applied parts coming, such as problem sessions and term project; the lectures will move in the latter part for more practical methods oriented direction; and there is still the recap and reading for the examination).


What is best in the course and what is the worst weakness? How would you develop the course in the future? Any other requests, questions or comments?

Some points from the written comments:

Some comments:

The feedback was good and we'll try to take it into account. Please find some miscellaneous comments below.

Text book. One of the main points in the comments is the text book and its relation to the lectures and problem sessions. The text book remains to be a recommended companion to the course.

However, many are studying the course without the text book, which may cause difficulties. In the planning of the course we decided to pick a good text book and we have tried to follow it, even if we would otherwise have preferred to treat some things differently. An alternative would have been to use several sources, which however would not have made anything easier. The lectures, problem sessions, term project, the text book and examination support each other. Of these, only the examination and the term project are required to pass the course. On lectures, one presents an overview and goes more in-depth in some issues, problem sessions apply the knowledge and text-book can be used for self-study, as a reference and source of additional information in the other parts of the course. Term project gives a more real-life application example. Examination hopefully motivates to participate to lectures and problem sessions and read the text book and other material. The course is designed to be purposedly modular so that the student can participate to the different parts in a way that best fits his or her study preferences.

We don't have written lecture notes (or scribes); slides are meant to be a representational aid for the lectures, not a replacement for the text book (in any case, an alternative set of slides can be found at the Alpaydin's site). We have assumed that the students have an access to the text book (at least to the reading room copy, there are two reading room copies in the TKK library, although we know that the library resources are severely limited).

We'll try to get some more reading room copies of the text book ordered. In the future courses one alternative is to use scribes, that is, some students would prepare summaries of the lectures as a course requirement or as an optional part of the course. Yet another option for the future courses would be use lectures for general overviews, or more detailed discussion of selected topics, while the students would in groups or independently study the topics in more detail from a given source material. Decent open sourced lecture notes would be ideal; this process would probably however take several iterations (as a side note, it would be really nice if there would exist and open source text book that everyone could use and improve upon; hopefully this will be true some day).

This said, it is the responsibility of the student to acquire the study material needed in the course. University studies require reading some books. Sometimes it is necessary to purchase book or at least copy some relevant chapters.

On topics. In addition to the availability of the text book, another issue is how closely the course follows the text book. We have covered some topics much more thoroughly than the text book (for example, Bayesian networks), some topics discussed in the text book we have skipped. On the other hand, some topics follow the text book quite closely, even though there is no time to go through everything in full detail during the lectures. There are also some planned part related to algorithms that are not based on the text book at all, but on a PDF chapter to be distributed from the course web site.

Motivating examples. The examples are not that real life in the lectures. It is partially a conscious choice, because to get the essential of the concept of methods one has to use simple examples. It is however true that one could additionally have more real life examples.

Lectures. About the lectures, some commented that they could be more clear, especially the use of blackboard was criticized. We'll try to improve clarity. (However, with respect to the blackboard, at least in principle mathematical topics are somewhat troublesome for presentation using slides; the blackboard is often better, when properly used. Lectures combined with a text book would be a good combination.)

Problem sessions. Several wished for at least some kind of models answers for the problem sessions, and said that without written answers it is difficult to follow the solutions, all of which are not represented completely or in full detail in the problem sessions. While we will not give away model answers to all problems, we'll try to arrange some written answers as well as more example codes. The course web site actually already includes some code, for example, most of the R code used to prepare the lectures or problem sessions (a known weakness in the current curriculum is a lack of proper computational methods course for data analysis). It should be noted that in addition to the Alpaydin's book other text books contain worked out examples, for example, Bishop's new book even has some problems for which the solution is available from the web.

Problems. Ambiguity of some problems is often more a feature than a bug (although sometimes mis-intentional one). Real world problems are not that well-defined either; often understanding the problem is more important (and difficult) than solving one.

We'll try to reward the problems properly in the grading.

You are at: CISTeachingT-61.30502007 → Mid-course survey

Page maintained by t613050@james.hut.fi, last updated Saturday, 27-Oct-2007 11:58:34 EEST