Sunday, 20 November 2005

Object virgin?

Recently, I have mostly been writing up some lecture notes on objects and classes. Now, an object is really just a bunch of data and some operations to manipulate that data. So, lists and operations to add and remove elements from them, whatever.

So, how to explain that to first year programmers? Well, I suggest not by introducing a "Pizza making robot" or a "Person" object. These are not entities which easily reduce to the "data and operations on it" model. I don't know what a person is, but I'm sure it's not just a collection of data and algorithms, even in a very abstract sense. As John says, a Person object really ought to have methods called "respire()" and "excrete()", but the ones in text books usually just have lame things like "get_age()".

In the end, students seem to respond better to programmatic examples (even if they are quite abstract) than really fake examples that are supposed to be "relevant" to them. Better to give them a PrintServer than a Person.

Tuesday, 15 November 2005

How to pass your final year thesis project

I've been thinking for a while that it would be good to distill some of the advice that colleagues and I give to students doing final year thesis project, so here it is -- enjoy!

Think of yourself as an academic.

Whatever you want to do when you leave University, your work will be written for academics and marked by them. A thesis project is like a small research project and to get the best marks available you should run your project with that in mind. So, you need to perform a thorough literature survey, follow a sound methodology, critically analyse your results and so on.

Have a hypothesis (or a thesis statement or a research question).

A lot of students approach their projects by saying "I want to do this", like "I want to invent a Foo, written in VB". This is not the best way to get good marks. Your project needs to have some sort of clear purpose and in the academic world it's best to state that purpose as a hypothesis, thesis statement or research question. These three are really just different ways of stating the same thing and which you choose is just a matter of taste. So, if your project is a study on the ratio of smokers which develop cancer (e.g. "I want to do some statistical analysis on smoking") then you might phrase that in one of these ways:
  • Smoking is correlated with cancer -- hypothesis
  • Smoking is correlated with cancer -- thesis statement
  • Is smoking correlated with cancer? -- research question
Or, perhaps you want to design a new GUI widget to replace list boxes:
  • Users will find WidgetFoo easier to use than ListBoxes -- hypothesis
  • Users will find WidgetFoo easier to use than ListBoxes -- thesis statement
  • Is WidgetFoo easier to use than a ListBox? -- research question
Part of the point of phrasing your work like this is that it should change the way you think about the work. You now have a very clear goal to reach -- to prove your hypothesis / thesis statement or to answer your research question. Everything you do in your project should now be directed towards this one goal.

Also, you have reduced your risk of failure. What if WidgetFoo turns out to be rubbish? Well, then you have still answered the research question or disproved the hypothesis -- you've got a result. In your write-up you can probably say a lot about why WidgetFoo wasn't as good as you thought, how you achieved your results and how other researchers can use your work to invent even better widgets.

Produce something useful to others.

So, you've got a hypothesis to answer. Great. But your teachers will still want to be convinced that the hypothesis you've chosen is worth six months of your time and lots of hours of their time. How do you know if your work is useful? Firstly, make sure you've read the relevant literature. Use Google, go to the library, talk to potential users. There should be a group of people in the world who have a clear reason to be interested in what you're doing. That might be a section of the research community, a user group, a company, whoever -- but there must be someone who wants your hypothesis validated or disproved. You should convince whoever's marking your thesis that this group of people exists by explaining in your thesis what the current literature says about the area you are working in and how your work contributes to this area.

Do something (a bit) novel.

You're not expected to win a Nobel Prize on the back of your undergraduate work. However, there's no point in doing something that everyone has done a million times before. The worst example of this in Computer Science is a project which is basically "I'm going to write a website with a database". Usually the website is for a friend or relative and the database keeps track of users. There's nothing wrong with that if there's some real novelty in it (e.g. you've just invented a new sort of database and this is a demonstrator for it). But often this is something that most first years could complete for a coursework (so, it isn't stretching you) there are simple, free tools that can do the job for you (it's not novel) and the user is bogus (noone's interested in it). Choose wisely!

Have clear success / failure criteria.

This should be taken care of when you write your hypothesis (or thesis statement or research question). However, it's important to know what will constitute a "successful" project for you. What results do you want to end up with? Once you know that, you can choose the appropriate methods to use in your work (e.g. will you be running a focus group? Writing a questionnaire? Writing some programs -- and testing them somehow?). You can also think about reducing the risks in your project. What if you don't finish part of your work on time? Is there some catch-up time in your schedule? What if early tests go badly? Is there time to repeat them?

Don't change the world.

Don't choose a project which is just far too big. If you're going to invent a new computer, writing an OS from scratch and a windowing environment for it is not a one-person, six-month, part-time job. Keep it small and give yourself some scope for extending the project if it goes better than you thought.

Even if you've chosen something small with a clear hypothesis, once you've written out a project schedule it will still feel like far too much to do. You will feel overwhelmed. The trick to reducing your fear early on (and making sure you work to schedule) is to break down your work into small tasks. Each task should have it's own goal and deliverables with it's own success criteria and a deadline. So, if your hypothesis is "WidgetFoo is easier to use than ListBoxes", your sub-goals might be:

  • Produce a literature survey on list widgets.
  • Write comprehensive unit tests for WidgetFoo in the Gtk widget set.
  • Implement a WidgetFoo in the Gtk widget set.
  • Debug (goal is to pass all unit tests).
  • Devise usability experiments (deliverable: methodology document).
  • Write application software for testing in the experiments.
  • Perform experiments.
  • Analyse data.
  • Write-up thesis.
Note that these tasks are dependent on one another. WidgetFoo cannot be written before you know how to test it. The experiments cannot be run until you have designed them and decided how to analyse the data they will produce.

Your new list will make life a lot easier, but you will probably still feel that it's too much to handle. To feel better about your work and motivate yourself, it's a good idea to take each of your sub-goals and write out the next physical action you need to perform to carry out that task. So, for the literature survey the next action might be "Google for 'list widget'". For the unit testing your next action might be "Find documentation about the UnitFoo testing framework". And so on. Most people find action lists much more motivating than task lists. Scroll down to see more stuff on productivity and where to keep your lists!

Choose a project that will maintain your interest.

Six months is a long time. If you choose a boring project (maybe you think it'll be "easy") then you'll quickly lose interest, get bored, stop working and possibly fail. In some ways, it's good to choose a project with a lot of scope (so you can change direction a bit and still address your hypothesis) in an active area of research (so there's lots of work to build on). On the other hand, some people would find that sort of project unfocused and confusing.

When you plan your project, think hard about what motivates you. Is it reaching towards a really interesting goal, or the fear of failing really badly? Either way (or both) you need to keep that motivation in mind whenever you're working. So, choose something interesting to do, give yourself at least a little scope for changing the details of the project over the year and keep in mind why you want to succeed (or not fail!).

Don't be too dependent on clients.

If you have a client to work for, especially one from industry, be careful about how you plan your project. If you are expecting resources from your client what will you do if they don't turn up on time? If you want the client to test your work, what do you do if they get a major order in when you're ready to test and noone has the time to help you out? What if the client goes out of business? What if you have to sign a non-disclosure agreement -- can the University still mark your work?

Having a "real" industrial client can be a big bonus. For one thing it's very easy to say that someone is really interested in what you're doing for the project. However, you need to make sure that if the client pulls out or doesn't cooperate you'll still have a viable project. Plan well and you won't have any problems.

Understand that research is not reading and a thesis is not a report.

Most undergraduates (at least in Computer Science) seem to be pretty confused about what research really is. It certainly isn't about using Google or reading in the library. Research means adding something new to the body of knowledge on a particular subject. This is why it's so important to know what work has already been done (so you know your work is novel), to have a clear hypothesis (so you know what new understanding you're adding) and to write up your work well (so other researchers can use it).

Also, understand the place of your thesis. You are not writing a report which tells people what you did. You are writing a thesis which tells people about the research you have done. This can be structured in whatever sensible way you prefer, but it needs to have the following parts:

  • An introduction. What's your hypothesis? Why is your work interesting? What are your trying to achieve?
  • A literature survey. What have other people done? What new knowledge will your work add? What is the current state of the art missing and how are you going to address that?
  • Your methodology. How did you go about validating / disproving your hypothesis? Why is your method sound? Why should anyone trust your results?
  • Your results. What did you do? How?
  • Your analysis of your results. What do your results mean? Why are they interesting? Did you validate your hypothesis or disprove it?
  • Conclusions. What did your work contribute and how could it be continued by others?

Eat well, sleep well, get some exercise and take a day off every week.

Basically, look after yourself. To work productively you need to be in good physical and mental condition. If you feel ill or your not coping well with life, slow down a bit and take a break. Don't eat junk food all the time or you'll feel sleepy and miserable. Cut down on caffeine and alcohol or you'll get stressed and sleep badly. Sleep a sensible number of hours every night -- consistency is important. Get some exercise because you need endorphins to keep you happy and some oxygen getting to your brain. Omega-3 and the sorts of minerals that aren't found in hot dogs will keep your brain working sensibly.

Most of all, take a whole day off University work every single week. It doesn't matter what you do with that day (have fun, earn money, write a novel, whatever) but working every day will limit your creativity massively. Most very creative (and productive) people find that their best ideas come after a day off. This gives your mind a chance to consolidate all the material you have learned, synthesise it and solve some of the problems you've been considering. In fact, to revise for an exam or solve an interesting problem, it's a good idea to spend a few days working really hard at reading everything you need to know and taking notes, then take a day off directly before the exam or the day before you're going to write the solution to your problem. This will give you the best chance to properly understand everything you're working and be creative about it.

Be productive but don't spend time on productivity.

You need to organise yourself well, which is a difficult problem in itself. However, if you spend even five per cent of your time on productivity management (e.g. using Microsoft Project!) then that is far, far too much and a massive waste of your most important resource -- your time.

Good productivity strategies are effortless, effective and fun to use -- and take almost no time at all. I can recommend David Allen's Getting Things Done strategy and RememberTheMilk for managing lists of action items.

One thing you can do to really give yourself a head-start in the workplace is to estimate how much time it'll take you to do every single task in your project. You'll start out finding that your estimates are stupidly far out, but as your project progresses you'll get better and better at correctly estimating your tasks. Joel Spolsky has a nice essay on how to do this simply, which you can use with RememberTheMilk.

Vote for this article on!

Monday, 7 November 2005

Python and stuff

Lately, I have been messing with the Python compiler module. It's a *very* nice interface, as far as OO things go and definately my second choice to SML.

On the teaching front, it seems that Adam is also blogging. No complaints about my lectures so far, but then I didn't check the archives...