Course reflection

Here I will critically reflect on what I have learnt in this course..

Difficult decisions in my poster design

The hardest things here were deciding what to leave out without losing the main points and how much time to spend on presentational aspects i.e. you can tinker with colurs and backgrounds for ever and not progress the actual content at all. Probably better to make an initial decision and then refuse to alter it until all the content is in.

Main differences between talking about my work at the poster session and giving a talk on it

The main diffierence was that during the poster session I was respondingto questions of people who had read it not proactively giving out information. In this sense it was more like the QA session at the end of the talk. If someone had come and said “OK take me through your work” then it would have been more like the presentation but using the poster instead of slides.

Mix of skills needed

Verbal, visual and written communication

What I did well

I think the poster and the talk went well, The slides were harder to create than i thought – they never seem to reflect what you actually say when you reach that slide, its like another person wrote them. The abstract took absolutely ages longer than i tought and opnions on it were hugely varied.

What I want to work on

it would probably be best for me to give a lot more talks as I’m on a roll and that is almost always my major weakness, either being nervous or appearing more nervous than I am.

Papers 2 & 3 – Neurogenesis and synaptogenesis

Sung-Bae Cho  & Geum-Beom Song  - Evolving CAM-Brain to control a mobile robot

The idea of using trying to grow a useful network by use of local rules only is very interesting and has a clear parallel in nature. The use of a cellular automata is clever as this is a very clean expression of growing complexity through local rules alone, there can be no cheating.

Clever solution to the problem of continuous analogue input into binary state cells. Using multiple cells of different sensitivity probably mirrors nature’s own solution.

On the downside, unclear if this could scale up to more complex behaviour. What is reliability of re-growing an effective solution from the same parameters?

 

Wensch & Sommeijer – Parallel simulation of axon growth in the nervous system

I like the way they have made a very strong and detailed model based case for the role of the chemical agents secreted and diffusing causing the kind of axon structures seen in nature.

Downside – would be nice to see some photos of real axon bundles to compare results visually. The simulation is incredibly detailed, it would have been interesting to start eliminating some factors to see where breakdown occurs – how many of these are actually necessary for the behaviour/ morphology seen. It would also be useful to reduce the resolution of the computations, how small is the critical homogenous area “units” for computation. Is it a process that can only work by very small localised precision? The level of detail and computational expense associated seems to have prevented any evolution based exploration of the parameters.

Paper 1of 3 – Neurogenesis and synaptogenesis

J. C. Astor & C. Adami – A Developmental Model for the Evolution of Artificial Neural Networks

The model of the aligning hexagonal occupy-able “spaces” is an effective and understandable simplification. I like the attempt to model the many factors during neural development including simultaneous cell and connection (axon/dendrite) growth.  A powerful feature is the inclusion of gene expression modification during the development/growth, producing cells that have the same genes but are producing different proteins and levels of each. This is a true dynamic simulation.

Also identifying low-level core network outcomes (self-limitation, logic gate development etc) is a powerful and useful analysis.

On the downside, the results section and conclusions are weak, I suspect because they have few because they have bitten off more than they could chew in terms of complexity.  By incorporating so many “genes” and cross-dependent factors they have made the use of the clearly advantageous GA approach difficult because of the search space size. This is compounded by what appears to a belief that the grown network must be judged on fitness of performing more complex higher level behaviours.  There is evidence that this growth stage simply sets the scene with a semi-blank canvas structure that can then adapt through plasticity to develop higher level and cognitive behaviours.

If they had included less factors they might have achieved similar results without having to resort to this complicated distributed GA architecture distraction. The justification for including so many factors up front is not clear – a more incremental approach might have been more effective.

Thoughts on my first paper

Well, that wasn’t toooo hard. After 3 years at university of being told I was unable to write in a sufficiently academic style and insisting on stooping to pop science this comes as rather a pleasant surprise. It would be nice to believe I had “cracked it” but I suspect that it had rather a lot to do with the fact that my co – nay -  second author has one of the most poppy styles of writing I have come across – and frankly its amazing, but reassuring, that he gets away with it. He has a clear belief in clear and interesting content over obfuscating jargon. And that is an exception I can tell you. For example, he has changed my abstract to be more poppy!

I quote: 

“When evolving Artificial Neural Networks (ANNs) to perform multiple tasks, are there advantages in tackling this one bit at a time, through incremental evolution?”

“one bit at a time” ? Das ist poppy

Another piece of my own advice, which I failed to heed fully, and paid the price, as I knew that I would – is to write the piece IN THE TEMPLATE THEY WANT. And that means all of it not just most. Why so you may ask? Surely you can just past in your text into the template and it will all flow in and be lovely.  Not if it’s in columns (and it always is) and has diagrams (it will have).

Columns and diagrams are like cats and dogs – they just don’t get on and 1000 years of human evolution hasn’t changed that.

Actually, we seem to have gone backwards – the Ancient Egyptians had cracked laying out columns of little pictures 5000 years ago. Try pasting the walls of King Tut’s tomb into Word 2007 for Vista and see how far you get..

What makes a good presentation?

Firstly, I would say it is critical to remember that “to present”  is not an intransitive verb .  There is a an object to whom it is directed – i.e. you (and I don’t mean the audience in general). What I’m saying is that a presentation can be fascinating or boring because of your interest in, and understanding of, the subject matter.  This is probably the single most important issue colouring your experience of the presentation. So the same talk can be success or failure for different members of the same audience. For example, I attended a talk the other day on the development of fishes’ brain. The speaker was good and the audience seemed interested – but for me it was impenetrable as the academic level and jargon was above my understanding of the area.

I believe also that the same factors will influence heavily your experience as a speaker.  Namely your own interest and enthusiasm and knowledge of the subject will make your talk far easier and more enjoyable to give, you actively WANT to impart this knowledge.  This in turn, you can be sure, will transmit itself to your audience.

On more specifics, one thing I’ve come to appreciate more here is the huge value of having something to break up the monotony of slides.  As an audience member these moments are often a balm to the drifting mind. In the past this had to involve either a demonstration or some kind of audience participation “Can anyone tell me..”. This latter technique is universally unpopular with audiences although it serves its distracting purpose. The former is universally popular with the audience but not with the speaker as it is often doomed to failure, usually in spite of extensive preparation. Indeed it is often the sight of the struggling demonstrator that is most succulent .  Now however a jewel has appeared on the horizon that solves all these problems at a sweep. Let us sing its name with great praise – It the embedded video clip.

Online Publications Again – Free to read?

Another huge issue for me has been whether a paper is available to read for free on the Internet or via one of the university’s subscriptions.   I find it baffling that the university is not subscribed to more journals (e.g. the IEEE ones which are very common in my field) but I barely need to ask the question why – the answer is obviously going to be “cash”. My wife has also suggested that I look for an inter-library loan but this doesn’t work for online publications.

Basically, if they want to charge for it then I am very unlikely to read it – however interesting it might look from the abstract. This is because they are damn expensive and may well not prove that useful, plus there are always zillions more to look. If my paper is published I would actively wish it to be freely available as this will certainly attract an order of magnitude more citations.

Again one has to wonder how long this will be sustainable. I think something has to give here – either they allow you to read part of the paper or else they become free or a lot cheaper – these are like $25 upwards.

My Peer Community

I am commenting on here on journals and conferences which are most relevant to my interests. Currently the journals I see most papers from are the Adaptive Systems journal and the Artificial Life journal . There also seem to be a lot of relevant papers published in conference proceedings e.g. the Alife conference which is the same one I have submitted my paper to. The IEEE conferences also crop up a lot but they tend to charge for these (more on this later).  Apart from those there seems to be a general mish mash. One of the points that comes home most strongly is that Internet is clearly making a huge difference to this – I’m sure that the days of the “prestigious” journal are numbered because the only thing driving this will be that is “considered” prestigious whereas before there was a clear driver – name the more important the journal, the more widely available it would be and the more widely read  – for example all libraries, not just universities would subscribe to Nature or Science. Now, so long as an article is available on Google Scholar then it does not matter which publication it appears in. At the moment, of course there is more competition for the “better” journals and a higher standard and stricter peer reviewing of submissions but one suspects that the days of this may be numbered.

Where’s (the) Wally

BTW there is a picture of me hidden in the header bar – see if you can spot it!

New Blog Header bar!

Created this by stiching photos of my studio. I like it!

Prolog project

Have had to face up to thinking about this. I think I’ve come up with something that actually uses Prolog as logic  – a semantic network. Need to be sure it will use recursion (yes) and lists (not sure, will have shoe-horn in somwhere) because there are marks attached to using these “amazing” features.