20 October 2008

Open University M366 - exam post mortem

The M366 results came in and I wrote a long post about it - but it disappeared. Great.

M366 is an Open University course, level 3, 30 points (should be more!) entitled "Natural and Artificial Intelligence". I managed to get a distinction, with something like 96 in the continual assessment and 92 in the exam. It was hard work, so I'm glad the paper had no nasty surprises.

So, for peeps thinking "Should I do M366"? Well - it depends. Are you interested and doing it for fun rather than aiming for a particular qualification? If so then yeah - have a go. Are you aiming for a degree and need a high mark to get the grade you want? If so then M366 might not be the best plan, the grade breakdown for our presentation (the first presentation) was:

y1pCsc9c3qMQ0Nk-GV4-gJTc9lcRRtcUFlmcM9cBx6STg78u_rG9L8WuMPGvXsp1Hlp

- so if you need a distinction there might be easier options. M366 is a lot more work than the other L3 I've taken so far (the concurrency one). M366 is multidisciplinary - you have a bit of maths (quite light), logic, computer science, programming, history, philosophy of AI. Some people come in to the course wanting to read about themes of consciousness, and philosophy. Others come in wanting to learn algorithms to help with their programs. Two different camps, both find parts of the course unfamiliar and difficult. The dropout rate was high.

If you do take M366 then...

  • Read ahead early on, it doesn't matter if you fail to grasp 50% of it, just read it all early on to get the gist of the underlying themes.
  • Allot more time to this than you usually do. I read the units twice before tackling the TMA, and again during the TMA.
  • Don't go nuts. I did, I spent ages perfecting my TMAs, going much further than required with the coding section. We had a TMA question that asked for a heuristic for a search and I knew what they expected within a couple of minutes, but I didn't like it as a solution and spent days making a better version. It wasn't really worth the time. My heuristic was faster and found shorter solutions though - much better (their one overestimated - a cardinal sin!).
  • Code some of it in <Insert you native programming language here>. I did some C# and VB.Net programs implementing search and genetic algorithms. It improved my understanding.
  • Unit 6 is critical for the exam. There is a compulsory question. The subject matter is not maths/programming and so it needed a lot of work. Write notes on stuff like the Chinese Room. Write an essay on the notes. Condense the essay down to something much shorter. Re-Write it the next day to reinforce your understanding. Do this for all the other junk in Unit 6. It will take a few days, but you will then be able to do the exam question easily.
  • Netlogo. Many folk detested this. Some were not programmers, and so they were dunked in at the deep end. To those people, don't panic. You needn't write any code - if you can explain in words how it should work then you'll get marks. Once you work out that, you might be able to takle the coding. Some were programmers, and they hated the language. NetLogo is from a different branch of the programming language tree to most of the common programming languages - C, Java, VB, LoLCode, etc. You'll see it is a LISP-a-like. It is worth finding a tutorial on LISP and having a go at the basic syntax (or SICP if you are a glutton for punishment). I liked NetLogo. Write some simple programs - get used to lists of things and making your own data structures from lists.
  • Rote learn. Create some lists for the "compare and contrast" type questions and rote learn them...
  • Don't leave TMAs until the last minute. You can't breeze through these ones as the Units are huge and complicated. I learnt miles more doing the TMAs than I did studying the units before doing the TMAs.
  • Leave at least a fortnight of late nights for revision.
  • Don't whinge if it is too hard. Take a look at the MIT courses...

15 October 2008

Open University - M362/M256 exams over

Exams finally over. I was a doofus and thought they were the coming Friday and Monday. I ended up with 4 nights for M362 and a weekend for M256 when I realised my mistake. I've no idea why, but I had the wrong date down on a bit of paper and I referred to the bit of paper whenever I wanted to work out how much longer I could procrastinate for.

Anyway, the both went ok in the end. I'll guess a Grade 1 pass for M256, and a borderline 1/2 for M362. It would be very nice indeed for that to be a 1, as then I can relax for my 3 remaining Level 3s. Otherwise I'll need to concentrate hard on one of them to get a Grade 1 pass - you need 60 points at Grade 1 in Level 3 courses for a first.

The past paper MCQ answers for M256, (Wednesday 17th October 2007 - M256/S) we reckoned to be:

1 2 3 4 5 6 7 8 9 10
AD BD DE BE CD AE BDF CE BC BF
11 12 13 14 15 16 17 18 19 20
BE CE C BD ABF BD BE AD CE C

5 - this is dumb, you read "an abstract class may have instances" and immediately think - "no way Hosé". But, in M256 parlance, they can have instances - these being instances of their concrete subclasses. Urgh. It's in the Glossary.

M362 (2008) was deemed fair by most people taking it. Some bombed and are expecting re-sits. I did Q13, which was a bold move. The paper is available now. I reckoned I would drop 5 or 6 marks on Q12, especially as I didn't like the last part one bit (security managers / RMI). So, I took Q13 - I liked the last bit on JAAS, but now I'm not so sure. I found the terms features and resources in parts a and c confusing - too vague. I guess it will be a grade 2. Widdle.

M256 (2008) was deemed unfair by most in the forum. A few people thought it went ok, I guess others are keeping quiet, not wanting to disturb the waters. The main complaints were the sheer volume of text you get for the main questions. They are pretty bulky, but most is just reference. Then folk didn't like the things you hand to copy out into your answer booklet and complete - a huge sequence diagram and a walk-through. I don't know if some people were using a ruler for their sequence diagrams. Don't, just scrawl it out. I ran out of room for mine. If your reading this with M256 approaching, then practice drawing a huge sequence diagram out on A4 - landscape, avoiding the edges (margins you can't write in). With 2 past papers available from the OU you should be in a better position to revise. I had a problem with Q23, I think there was a method missing in the protocol - so I had to make one up - a getter for something or other. I'll check when the paper comes out. I couldn't see another way of getting the information (alternative navigation of links, or cunning use of other methods), and the walk-through seemed to indicate that the thing was available without having to go via another object.