M.A. (Cantab), M.Sc., Ph.D.
The Computer Science Era
Aston
My computer science period commenced when I did an M.Sc. conversion course to
Information Technology at Aston
University in Birmingham. There my project was on Genetic Programming, tying
together my first two degrees.
As the amount of program code that has to be written rises ever further, it
would be advantageous for computers to be able to program themselves. (Already
today most of the code written is generated automatically.) Genetic Programming
is an attempt to use the mechanism of natural selection to evolve computer
programs. To an initial primordial ooze consisting of a population of randomly
generated programs, one applies the genetic operators of crossover
(recombination), reproduction and mutation, selecting the programs according to
their fitness for the task in hand. Over the course of many generations
evolution takes place to give better programs; and hopefully they eventually
manage to solve your problem. GP has been used in certain fields to give
comparable results to humans; and in a few cases has given better results than
human-constructed solutions.
At Aston I implemented John
Koza's original GP system and used it to compare the performance of
standard GP, GP with Automatically Defined Functions, Conor Ryan's Pygmy Algorithm,
Perkis' Stack GP; and various combinations of these with each other. (This
description will also be expanded when I find time.) I also used GP to solve a
problem of my own; namely the generation of mazes.
My M.Sc. thesis is now available on-line.
Heriot-Watt
After graduating from Aston with a distinction, I went on to do a Ph.D. at Heriot-Watt University in Edinburgh, on
sensor planning for Machine
Vision. This consists of, for example, ensuring the camera is pointing at
the objects you wish to examine, making sure the features to be identified are
not occluded or around the back of the object; making sure the objects are not
too large or too small in the field of view, and are in focus, and making sure
the lighting is appropriate.
My work was to see whether the problem of sensor planning was amenable to
solution using Genetic
Programming, thus building on the knowledge from my M.Sc. project. My Ph.D. thesis is available here in zipped PDF (3Mb).
My Ph.D. work was done in C++ and Tcl/Tk.
Once released, the colicin can only operate once it has got into other
bacterial cells. To do this it must pass through the bacterial membranes;
which it does via a mysterious process involving binding to the
porin proteins
on the outer membrane. This process can be demystified by observing the
colicin as it binds to the porin, by means of fluorescence spectroscopy.
However if either colicin or porin contain any aromatic tryptophan amino acid
residues, they will fluoresce at the same wavelength and swamp out the signal
you are looking for. Consequently my project was to engineer a version of the
colicin protein that had no tryptophan residues but was otherwise structurally
and functionally unaltered.
,
RTS nominated 
) -- the world's first interactive TV adventure, in which the
decisions you the viewer made via your remote control determined how many
clues the treasure hunter got, and ultimately whether he found the treasure
or not. Later services of a similar nature included Death In Rome (a
murder mystery: was Claudius Eutychus' death natural, and if not, who
killed him?), and Spooks interactive
(
BAFTA winning, winner, Best Game & Entertainment and Jury
Grand Prize at the 3rd AFDESI International ITV Awards).
BAFTA nominated, 
EMMA winning)
(website), in which
you played along with the TV series on your remote control, and at the end
the set-top box would mark you and give you a score breakdown, and (though
not on digital terrestrial) forward your result back to the BBC, along with
players on the web, so the presenters could give out statistics as to how
the nation as a whole was faring.
), with extra video streams,
medals tables, schedule of events and news stories.
![[monogram]](images/monogram.png){kind=small}
