Unsuitable Project Topics

The Examiners are concerned that some of the topics chosen for
Part II and Diploma projects are unsuitable for Cambridge
students, and unlikely to lead to a satisfactory dissertation. In
2005 we considered the 20 projects ranked lowest in the final
results, and have identified common features of some unsuitable
projects. We advise against these projects being chosen in future
years, and describe the attributes that in our opinion led to the
problem.

One of the most important attributes for a successful Tripos or
Diploma project is already well known by overseers and Directors
of Studies, and is emphasised to students in briefing material.
This is that it must be possible to clearly identify two phases
in the technical work: firstly a core element of the project that
is definitely achievable, and that would, in itself, form the
basis of a good dissertation; and secondly an extension element
that is likely to be within the abilities of a Cambridge
undergraduate. The projects that we have identified as unsuitable
choices failed to satisfy one or both of these criteria. Examples
follow:

Applying genetic algorithms to the game of Go
There is no satisfactory partial solution to this problem. Simply
encoding the rules of Go is not a sufficiently challenging
project as a core project. The extension problem, creating an
adequate computer Go player, is well known to be a research
problem that has frustrated professional researchers for many
years, so is unlikely to be within reach of an undergraduate.

Advanced fee fraud email analysis
The core part of the project either assumes changes to basic
Internet protocols, so could not realistically be deployed on a
scale suitable for testing, or is a simple dataset reading
exercise. No suitable dataset was available. A successful
extension would constitute automated prediction of large-scale
social behaviour in a manner that has not yet resulted in
successful research results.

Joystick control of motorised stage
The core part of the project is a basic electronic interfacing
problem, and insufficiently challenging. No extension was
identified that would bring any part of the Tripos syllabus to
bear.

Distributed trust management
There are no metrics by which the core part of the project could
be evaluated for success without undertaking a huge user study.
No suitable dataset is available, and obtaining one would involve
large-scale technical work. Once again, a successful extension
would require prediction of large-scale social behaviour.

Microsoft Word document cleaner
Core part requires reverse engineering of a proprietary format.
Even if fully successful, there is little of interest that can be
said in the dissertation beyond the simple fact that it was
achieved. 

Heartbeat pattern analysis (and other projects involving corpus analysis)
The real interest in such projects is in the data, rather than
the software built to analyse it. The core elements include
reading the data format in which the corpus is obtained, and
implementing basic analysis algorithms, neither of which is
sufficiently challenging. Novel analysis results may represent a
research contribution in some other field (medicine, for
example), but will not attract additional credit for the
dissertation.

Context aware media player
The core element, constructing an MP3 player, can be easily
achieved using standard Java libraries and/or many open source
implementations. The extension to automatically anticipate user
preferences has been an active research topic with little
progress over many years. It is unlikely that an undergraduate
would make any progress without large scale user studies,
substantial innovation in HCI theory or a novel technical
approach, none of which had been identified at the outset.

Recommendation

It is the Director of Studies who is in the best position to
assess whether a project is suitable for a particular student,
although project overseers are able to warn both the student and
Director of Studies when the topic itself, or the way in which it
is approached, seem likely to lead to trouble. In at least one
case this year, overseers strongly advised against a choice of
topic, but the student and DOS chose not to take this advice. The
resulting dissertation was awarded nearly the lowest mark this
year. The Examiners in 2005 believe that choice of topic is
critical to the success of projects, and strongly recommend that
overseers and DOSes do their best to warn students away from
topics like those above. If students persist with unsuitable
topics, they are very likely to receive a dissertation mark near
the bottom of the class.

Alan Blackwell
Steven Hand
David Greaves
(CS Tripos and Diploma Examiners 2005)