Well, I can either spend time doing it or spend time telling you about it, which would you think a better use of public funds? And which would I be likely to find more fun? Right. Then don't expect too much.
Most recent PaperFor a look at the way the game is played, there was a warts and all stage which goes back to January 1997 and where the documentary evidence has been completely lost. There was some small amount of mathematics in it, mostly geometry with some hints of analysis. After an extended period of illness I returned to this paper and finished it off. The next version was submitted to TPAMI, the IEEE Transactions on Pattern Analysis and Machine Intelligence on April 1st 2004 under the title ‘The Geometry of Video Images’. Here it is as a pdf file. The date of this version is May 2004. There were two reviews; neither understood what the paper was about. One commented adversely on my English, complaining I had too many subordinate clauses, a complaint which would have had more force had the referee been able to spell the word `declaratory' correctly. He didn't mention the content or the mathematics. The other referee tried harder, showed some grasp of what it was about, but failed to understand the whole point, which is that if you know some data came from a smooth manifold with noise you can at least try to perform some kind of local regression. This process is called "Peer Review". It appears to be designed to slow down any new ideas. The second referee's main contribution was to suggest that it be sent to the Journal of Mathematical Imaging and Vision. I did this and after some more referees comments, the final paper was published in November 2006. You can get it online if you have a friendly university library subscription by going to the Springer Page and following the online links to the November 2006 issue. You can't get it from me because Springer owns the copyright. During this period I was, you will be pleased to hear, doing other things. |
My primary interest is in modelling cognitive processes in a way which is (a) compatible with what is known of the neurophysiology of the Central Nervous System and (b) consistent with some of what is surmised by psychologists. In fact the limitations of what is known to either group are considerable. Much is known about the behavior of a single neurone, the Huxley-Hodgkin model is well understood, and has stood the test of time fairly well (See May 1996 Spectrum). The physiology is reasonably consistent with the model, and more details are being added daily. The situation is not so satisfactory for ensembles of neurones. It is known that there are edge detecting neurones in the visual cortex and that layering implies that there may be constellations of edges gathered into more complex arrangements. It is also known that small children can infer grammatical structure at an early age. These two elements may be put together in a model of generalised grammatical inference which is done in the draft paper on cognitive processing given below. The paper is at present in the scruffy, pre-draft form of the paper on Lie group transformation of images at the top of this page and has a long way to go before it can be sent to a journal. Note that this draft is an accumulation of many years joint work with Dr. Chris deSilva.
The following pdf file contains some preliminary material on syntactic filtering which should give a painless introduction to the ideas I am currently working on. At a somewhat more demanding level, the pdf file shows where I am at with an early, draft, very scruffy stage of a paper on modelling cognitive processing.
About three weeks later, the above paper has grown and now looks like the June version. I must point out that this is currently in the first pass state, which means that it is scruffy, tendentious, contains lots of unsupported claims and is generally far from being a respectable paper. This is how they start life. Respectability comes later. If at all.
It is now 8th August and the first pass is finished. It can only improve. I shudder when I read it, it is flaky and full of holes, and riddled with stream of consciousness, showing all too clearly how it came to be. Of course, this might be a fascinating historical document one day. There again, …
The paper is called `The Mathematical Principles of Cognitive Processing’. On good days it looks to me to be about as important as special relativity, and the title moderately descriptive. On ordinary days it looks wildly hubristic, and on bad days it looks to be an exercise in special pleading. On the other hand, writing it is addictive.
I have called this version `fatcogproc’ to distinguish it from the thincogproc which will emerge when I have replaced all the verbiage with algebra. This will add respectability by contemporary academic standards but reduce intelligibility, but so many reviewers seem to feel that if they can understand it, then the paper can’t be any good.
Now comes the task of making it more respectable. This means organising it more coherently and leaving out the chat.
I have to say that for most of the past two months, working on the damned thing has been like clawing very small diamonds out of quartz with your teeth and fingernails. Not that diamonds come in quartz, or that it is easy to claw with your teeth. Nor, for that matter that the material is diamond in either brilliance or value. I am thinking of taking a week off and getting some exercise.
There was a long gap of time during which I did other things, some involving teaching myself differential geometry so I could understand more Physics, and some confronting the fact that a 63 page paper is unlikely to get published. So I started to turn it into a book which you can struggle through if you are brave. It is unfinished but I shall get back to it when I have time spared from students and the distractions of the Mike and Ike show.. Not to mention the distractions of updating this web page. I should say that the only reason it isn't a lot further out of date than it was a few days ago is because I wrote an article to be published in Quadrant, June 2007, a magazine for intelligent people. And the editor wanted people to see my website, so here I am in May 2007 making it more or less contemporary. If you read this in 2010 it won't have been touched for three years.
Any comments or suggestions of a
constructive nature welcomed.
mike alder, May16th 2007