Dates, etc., are correct, I think, but I wouldn't wish to guarantee all of them. Titles given are not necessarily those of the documents recording the work - they are descriptive of the topic. In most cases, I have some record of their work, either as thesis or report. Ask me if you're interested.
These entries are in chronological order of enrolment, more or less, then alphabetical order by surname, with occasional vagaries as I decide whether or not to sort by qualification. If you know for whom you're looking, you might like to try the index.
Links connect the set of entries for a student, and the set of entries for a topic. They're circular, so you'll get back to where you started if you persevere.
If you are honoured ( if that's the word ) by being included in this list, and would like to correct or extend your entry, do let me know. Observe, though, that the descriptions are presented from my point of view, and are designed to show how your work fits in with other related work on the same general topic. I might therefore emphasise some component of your work which you didn't think was very useful, or vice versa.
Some of the entries are sketchy, others are presented in more detail. If that means anything at all, which is questionable, it's that I've wanted to fill out topics which have led on to something else, or which I thought were going to lead on to something else but didn't. Whatever, it isn't personal.
Well, Paul was a chemist, so I lied, but he was working on a computing topic. Unfortunately, he didn't finish; he was lured away by the offer of a lectureship at Massey University, where he still is. Paul got as far as an implementation of something like Eccles on the B6700 before leaving, but I never had time to carry on with his version.
More Eccles ?
I didn't supervise Reg; he was a student in the archaeology section of the Anthropology Department who approached me for advice while I was in the computer centre. Reg was trying to estimate the number of fish which must have contributed to piles of fish bones excavated from old Maori middens, so that he could guess at the size of the population. We worked out a method which was probably a variant of dynamic programming.
More dynamic programming ?
Some further information :
I think I co-supervised Mary's work; Bruce Biggs was the other supervisor.
Mary used a method rather like that I worked out for Reg Nichol's fish bones ( see above ) to search through a database of words in various Polynesian languages looking for cognates - related words with similar derivations. We hypothesise that dynamic programming is the unifying principle of Anthropology; we can't think of any other.
More dynamic programming ?
Some further information :
George Blanchard was Elaine's supervisor, but her work was part of the PFL development. I'm not sure now whether or not I was an official cosupervisor, but I was there somewhere.
Elaine developed a virtual machine which could, in principle, run the code produced by Peter Sergent's compiler. The machine worked satisfactorily on synthetic test programmes, but I'm not sure that it ever ran with a compiled file.
Some further information :
More PFL ?
I shared Peter's supervision with ( first ) Tony Gear of the Management Studies department ( interested because Peter's proposal mentioned simulation ), and, later, George Blanchard of the Mechanical Engineering Department ( who would have been in from the beginning if he hadn't been on leave when I was looking for a cosupervisor ).
Peter was the first person to work on PFL. ( The initials originally stood for "Peter's Famous Language", but we later respectabilised them to "Process Formulation Language". ) During his work, we designed the first version of PFL, and Peter wrote a compiler for it which produced code for a virtual machine.
More PFL ?
Some further information :
The possibility of using parameters with Cobol PERFORM instructions was discussed, and a simple implementation programmed and evaluated.
Some further information :
Robert designed a Forth-like language and implemented an interpreter for it, called RATS ( Robert's Amazing Threaded System ). He used the language to write programmes for a simulated robot.
Some further information :
Sean designed an operating system for a 68000 microcomputer system.
Some further information :
Martin worked on the development of a chemical line-formula notation.
Some further information :
Steven wrote a programme to play Lasca. It used a straightforward minimax algorithm with lookahead, and performed quite well..
More game playing ?
Some further information :
Umesh got some way with an implementation of Eccles. As he wasn't a chemist, this was not a direct successor to Paul Lyons's work, but the logical processor was the same. Umesh extended the Eccles idea to work with textual statements, which led to a sort of higher-order logic which we didn't characterise.
More Eccles ?
Some further information :
Robert's main concern was to assess various artificial intelligence techniques as conventional methods in the software armoury. In the process, to illustrate his ideas, he worked on a programme to advise students enrolling in university courses.
Some further information :
Shane was originally going to try to develop a robotic feeding machine, but the robot, supposed to be provided by an external organisation, never turned up. He did some work on the design of such a machine, but unfortunately never recorded it.
The simulator was one of the early attempts in the department to use the Macintosh graphics. It was slow, but worked.
Some further information ( about the original proposal ) :
Lex tried to write a programme which would accept a stream of real numbers ( such as a sampled signal ) and attempted to discern patterns in it. Though he had hoped to use data from sensors mounted on a boat, we never got them, so he used patterns from one of the Unix monitors. It worked.
Lex's work was the beginning of the adaptive interface development. ( Lex would be surprised to hear that, but the idea was lurking in my mind at the time. This was before my leave in 1988, which was when I convinced myself that rehabilitation computing was a topic justifying significant attention. )
More about the adaptive interface ?
Some further information :
Robert designed and developed a computer system with which a young man with a spinal injury could communicate with his mother for day-to-day requirements. The mother was deaf and couldn't read very well; the young man had been accustomed to communicate with her using a sign language, but after his injury was unable to use his hands. Robert's system made it possible for him to select a request from a menu and display a corresponding picture on the computer's screen.
Some further information :
More exploits of Robert ?
- and here we skip a year, because I was on leave in 1988.
John studied my implementation of the "append" operation in my version of Eccles, which I wrote while on leave in 1988. "Append" is used to disjoin a term with a disjunction of terms, and carries out a number of simplifications intended to prevent the composite term becoming more complicated if possible. John was able to demonstrate that "append" was properly implemented, and that it correctly executed some, though not all, possible simplifications.
More Eccles ?
More exploits of John ?
Some further information :
Graham studied the requirements of simulation software, and designed a simulator which would combine continuous and discrete simulation. Very unfortunately, he didn't have time to write the programme - I think it would have been very good.
Some further information :
Max was enrolled in the Management Studies and Labour Relations department, but, due to circumstances outside her control, was unable to complete her Ph.D. Max came to the Computer Science department in search of a technologist's view to balance the management view endemic in her home department, and was directed towards me in my capacity as he who talks to weirdos.
Max's major interest was in personality type differences between typical managers and typical data processing people. It was very educational for me and enjoyable for ( I hope ) both of us. I think it was very unfortunate that extraneous problems made it impossible for her to continue with her work.
Mark started a new trend. He'd taken my real-time computing course in 1989, and had been working with Fisher and Paykel ( F&P ) during the summer vacation. Some things he'd seen there engaged his interest sufficiently to encourage him to follow them up as a research topic, so he persuaded me on the one hand ( without much difficulty ) and F&P on the other hand ( with rather more difficulty ) to sort out a topic for him.
The intention was to investigate programming languages and techniques which would be appropriate for constructing software to be used with a production line. Special constraints were to be observed; the most important, in the long run, was that the programme had to be robust to changes in plant, manufacturing techniques, and product, such as are realistic expectations for a production line evolving over several years. Using conventional techniques, even a slight change in conditions or environment could necessitate a massive programming effort to bend the software, and as the rate of change accelerated with time this programming load was becoming intolerable.
This was the beginning of PDL. It was intended as a highly modular structure in which all essential details of the plant, process, and environment could be recorded and easily changed whenever circumstances so required, together with software which would generate the appropriate control programmes from this information. Any change could then be communicated to the system by rather simple database changes, whereupon the rest of the software would adapt itself automatically.
Mark discussed and analysed the problem in some detail, and implemented an approximation to an interpreter which would, to some degree, satisfy the constraints. He got far enough to present a convincing case for the effectiveness of this approach. ( Well, it convinced me. )
More PDL ?
Some further information :
Roy studied the possibility of using a computer to marshal arguments, recording the progress of a debate, and trying to keep it honest by identifying inconsistencies, points to be resolved, and so on. He didn't get very far ( I hadn't expected him to - I'd suggested the topic for an M.Sc. thesis, and hadn't expected a great deal from that ), but it was an interesting and worthwhile start.
More exploits of Roy ?
More Debate moderation ?
Some further information :
Peter Gibbons was Simon's real supervisor in this project, but I, officially the assessor, took an active part in the discussions and development, and I've included this entry because it's the direct precursor of Simon's M.A. work in the following year.
Simon investigated the use of authoring packages as means of developing software to be used by retarded children.
More exploits of Simon ?
Nick wanted to work on language understanding, and constructed a Unix advisory programme as a case study. He was able to implement a complete system, which could accept an English question, determine its meaning, match the question with data on the functions of Unix instructions, devise an answer to the question, and construct a solution in English.
It was mildly amusing that the system was able to answer questions about only one Unix instruction ! - but Nick found that to encode the required information about the instructions was a very difficult and time-consuming job, and he didn't have time to do any more.
Some further information :
John was keen to explore the potential of neural networks as tools for handling linguistic problems. He was particularly interested in the possibility of perceiving patterns in input streams, and of deriving from such observations the grammar which determined the syntax of the input.
John's project extended over two years. We knew it would be a big one when he started, and had hoped that he would be able to do the work as an 8-credit project over two years, but this fairly modest request beat the university's administration, so he had to take two 4-credit projects. He appears again below in 1991, but that's just a formal entry.
More exploits of John ?
Some further information :
( This entry is here under false pretences, as I was Bernie's assessor; Peter Fenwick was his supervisor. The pretences are only slightly false, as I was involved with the project throughout. )
Bernie analysed the performance of a commercial online transaction processing system, and made recommendations for error recovery procedures, covering topics including probability of hardware failure, software engineering principles, and human interface management.
Some further information :
Roy worked on a reaching device which we called the Helping Hand. ( Another device of the same name has unfortunately been produced commercially, but we didn't know that at the time. ) Our Helping Hand was a "robotic" reaching device operating as a bridge rather than the more common cantilever; we hoped that this form would be more rigid with cheaper and lighter machinery, thus leading to the development of cheaper reaching aids.
Roy produced a control system for the arm's motors, including the motors, the electronic hardware to drive it from a Macintosh, a design for a comprehensive bus for this and related control problems, and software to make it all work. Unfortunately, he didn't have time to produce the arm.
More exploits of Roy ?
Some further information :
Mike was originally going to work on a remotely-controlled vacuum cleaner for people unable to drive a conventional model, but just after he'd got under way some Korean firm came out with a commercial model. We thought we should change to something else in case the Koreans had done it all. ( In hindsight, I wish we'd gone on; Mike's work was such that I'm sure he'd have done better than any commercial model. )
Instead, Mike developed a domestic communications network which used a combination of house wiring and infra-red links to establish communication between many fixed or mobile units and a controller. It was more effective and faster than other devices available at the time, and it included facilities for units to respond to messages as well as simply react. I saw it as a way to find out more about domestic networks with particular applications to facilities for disabled people, but there was no opportunity to take this line of work any further.
Some further information :
I shared Simon's supervision with Keri Wilton of the Education Department. Simon approached the Computer Science department in search of anything to do with computers which might help him in his educational interests, was directed towards me in my capacity as he who talks to weirdos, and stayed. ( Part of the credit belongs to Peter Gibbons, who was Simon's first target as the lecturer in charge of our multimedia course at the time. )
Simon evaluated two authoring packages for their suitability as tools for schoolteachers.
Some further information :
More exploits of Simon ?
Hans visited us for a year from Kaiserslautern University ( Germany ). He was working on his Diplomarbeit, which corresponds reasonably well to our M.Sc. thesis. He worked on the adaptive interface topic, investigating the potential of neural networks as an implementation technique.
Hans's main contribution was to analyse the requirements of the project, and thence to demonstrate that neural network methods would be adequate and appropriate for much of the processing requirements.
More about the adaptive interface ?
Some further information :
( See John's 1990 entry. )
Adrian followed in the footsteps of Mark Brodsky continuing the development of industrial control systems, and particularly of PDL, though he thought it worth starting from scratch with some aspects of the design. Mark had concentrated on overall analysis and implementing the interpreter; Adrian leant more towards the database side of the system, and as a part of his work implemented both an interpreter and a model object-oriented database system. Once again, I was convinced that the approach showed promise, though F&P were less enthusiastic. We ( F&P and I ) parted ( I hope ) friends.
More PDL ?
Some further information :
Worik had intended to work jointly with another student on the development of a micromouse - a small autonomous robot which can find its way through a maze. ( There's an international micromouse contest every so often. ) Unfortunately for his plans, the other student dropped out almost immediately; Worik went on by himself, learnt a lot, but wasn't able to complete a model.
Some further information :
I shared Chris's supervision with Richard Lobb, and we received generous help in advice and cooperation from staff of the Kelston Deaf Education Centre.
Chris's intention was to investigate the feasibility of a computer system for drawing and animating signs and to construct a prototype. It was intended that the system should accept passages of plain English text and present these as Signed English on the computer screen.
The figure displayed was to be entirely synthesised, not merely retrieved as a sequence of digitised pictures from a library. This would make it possible, in principle at least, to present an approximation to a human signer's continuous motion, rather than a disconnected sequence of images. It would also be straightforward to alter the speed at which the signs were displayed, from a slow teaching speed to a faster more conversational speed. The highest achievable speed would depend on the characteristics of the machine being used.
As an additional advantage of the synthesis technique, it would be possible to view the figure from different angles and distances, where emphasis of the three dimensional nature of signs is required.
The job was bigger than we expected ( surprised ? ), but in an interesting way : Chris spent much time exploring possible ways to encode hand positions, and did a creditable job on it. He also managed to present a useable picture of a hand from his encoding method, but he never got as far as a moving presentation. Unfortunately, despite the link below, no one has ever wanted to carry on with this line of work.
More animated signing ?
Some further information :
Peter followed Roy Davies's 1990 project on the same topic, but didn't get a lot further.
More Debate moderation ?
Some further information :
Graham wrote a programme to play Lasca. It could be used at various levels of look-ahead, and its performance varied from blindingly fast but stupid to excruciatingly slow but really rather clever. He had hoped to experiment with different strategies, but unfortunately there wasn't enough time. He did achieve a very smooth interface.
More game playing ?
Some further information :
I shared Tim's supervision with George Coghill of the Electrical and Electronic Engineering Department.
Tim worked on a neural network technique for computer vision. His aim was to develop a method which could be used to recognise "naturally variable" objects sufficiently quickly and precisely for use as robot controllers on a production line. A "naturally variable" object is typically something which grows, so that each instance conformas to the same pattern even though no two may be identical. Tim used tree leaves in his experiments; my interest was in the possibility of using the same method for faces.
It worked, just about - and that was a very significant achievement.
Some further information :
I shared Brent's supervision with Mike Corballis of the Psychology Department.
Brent was using neural network methods to construct models which would illustrate or test Mike's theories about people's (in)ability to distinguish left from right. He did some very interesting work, but very unfortunately didn't write it up.
Kim was working on the development of a computer interface for people with severe physical disabilities. It is intended that the interface should be trainable, so that it can cater for a wide range of disabilities, requiring a correspondingly wide range of configurations of input devices. This adaptability is important so that a single system can be used for many people, and thereby gain the economic advantage of comparatively large-scale production. It also gives rise to many interesting computing problems.
Kim suspended his PhD work for personal reasons, and was not permitted ( because of university regulations ) to extend the suspension for long enough, so he withdrew from the PhD programme.
More about the adaptive interface ?
Paul worked on intelligent, primarily robotic, systems, which could learn about their environments by exploration, and then use the knowledge they had gained to accomplish arbitrary tasks. He argues that, as the complexity of tasks and real environments increases, it becomes impracticable to programme them by any form of conventional manual technique, so that an autonomous self-programming system is essential.
This led him through many stages of simplification and exploration, many of which are not ( so far as I know ) recorded. This is unfortunate, because his journey of simplification was a fascinating exercise in discovery, and exhibited an unusually high level of intellectual rigour and discipline. He finally endeavoured to embody his ideas in an implementation based on reinforcement learning, which was fascinating.
In the absence of visible documentation, I cannot do justice to his work. All I can say is that I did try to suggest that he should write things down more.
Paul wrote his thesis and it went through one round of assessment. He was asked to rewrite it, but finally decided against it. I think that's sad, but he had his reasons and it was his decision.
Some further, very scrappy, information :
Mark was primarily interested in the notion of modularity in neural networks, but as a part of his work he implemented a composite neural network system which did interesting things with computer vision. This was a very ambitious project, and he succeeded in most of what he was trying to accomplish. Most important from his point of view, he finished up with a system composed of many modules, which gave him good insights into his main topic, but he also implemented two particularly noteworthy techniques in his software.
The first of these was an edge-detection system using wavelet techniques. This was his own idea, and seemed to me to be very effective.
The second was the distortion network, a connectionist device to map images onto templates and thereby identify their components.
More distortion network ?
Some further information :
I shared Brian's supervision with Robert Sheehan ( more about Robert ? ) of the Computer Science Department.
Brian was interested in pursuing some sort of development in rehabilitation systems for his project, and, as he had extensive experience in giving introductory computing courses to blind students, thought that would be a good avenue to follow. We originally took up a suggestion from Massey University for work on machine production of mathematical or musical Braille, but discussions with people from the Royal New Zealand Foundation for the Blind convinced us that those topics were not good places to start. Instead, they suggested that a Braille tutor for sighted transcribers would be more use, and Brian worked on the spelling checker as a preliminary contribution to such a system.
Some further information :
Simon was interested in the use of multimedia techniques for teaching, with particular - though not exclusive - emphasis on the needs of mentally retarded pupils. Later he became interested in blind people's interaction with graphical computer interfaces. His literature work has shown that we know very little about what multimedia techniques really do to people, so that an area of understanding important in systematically devising multimedia presentations is absent.
Simon was unable to complete his doctorate because of his work commitments.
More exploits of Simon ?
I shared Roger's supervision with George Blanchard of the Mechanical Engineering Department.
Following Roger's notable work on PFL and PDL in my Robotics and Real-Time Control course ( which led to a technical report ), he embarked on the production of a PFL compiler for this project. And he wrote one, and it worked. On the way, he redesigned the language and produced a virtual machine to run the compiled code. That was a very impressive project.
More PFL ?
Some further information :
David investigated the development of cooperative processes in a one-dimensional cellular automaton, using a genetic algorithm to explore the space of possible rules. David had some very exciting ideas about such systems, and it was unfortunate that he ran out of time before he had thoroughly explored them.
Some further information :
Stephen had hoped to construct - or, at least, lay the foundation for - a system which could analyse the sounds made by deaf people as they try to speak and, by inferring the configuration of the vocal tract when the sound was made, offer effective advice on improving the intelligibility of the speech. Unfortunately, he didn't get very far with it.
I shared Igor's supervision with Clark Thomborson of the Computer Science Department; Hans Guesgen took part in the supervision earlier.
Igor was concerned with computer vision, for which he intended to exploit the properties of neural networks and fuzzy systems.
He achieved some success, but eventually withdrew because he ran out of time and money.
Tim's primary interest was in intelligent techniques for robot vision, but as the complexity of the problem became more clear he transferred his interest, for a while, to a study of modular networks.
He then did some work on the distortion network, and investigated implementation techniques.
It was unfortunate that Tim had to withdraw from his PhD research because of pressure of work.
Some further information :
More distortion network ?
I shared Doug's supervision with Dave Helweg of the Psychology Department.
Doug's work was concerned with reinforcement learning, but in his investigations he found himself led in many directions.
His original intention was to explore the potential of a neural network of his own design, which he called the CrushNet, but in practice this worked less well than he had expected. After a number of excursions, dominated by Solomonoff's ideas on the relationship between the simplicity of algorithms and the number of bits needed to describe them, he decided to investigate self-conscious machines.
He had some interesting ideas, but unfortunately never wrote up his thesis.
Fai worked on the automatic generation of helpful navigation instructions for blind people. The intention was that, given the plans of a building ( for example ), the system should be able to give clear and effective instructions for a journey between any two places in the building.
Some further information :
Duncan worked on the distortion network, starting where Mark Scaletti had left off, and got it going on a few simple diagrams. He was able to eliminate some complications introduced by Mark, and to demonstrate that the network would still work, but didn't have time to evaluate the method thoroughly.
More distortion network ?
Some further information :
I shared Natalie's supervision with George Blanchard of the Mechanical Engineering Department.
Natalie worked on programming for control systems, continuing the PDL saga. She implemented enough of a very simplified system to give us a fair bit of confidence that it was really going to work some day. In particular, she was able to experiment with the database organisation; her simplified system used three databases, and it all tied together nicely.
More PDL ?
More about Natalie ?
David evaluated the CIT3 system, which is a small hand-held computer operated through a touch screen.
Some further information :
- and I was on leave in 1996, too, but we don't skip a year this time. That's the miracle, or curse, of the internet.
I was Robert's sole supervisor until 2001 ( except for a period during which I was on leave ); then the university regulations changed, so Jennifer Lennon joined us as a cosupervisor.
Robert was ( and still is ) interested in the design and functions of computer systems for children; in particular, he is working on a specialised world-wide web browser which was once called "Scooter". Since then, it has passed through many incarnations and changed wonderfully with experience gained. It is now called ICE ( Incremental Computing Environment ), and about to suffer yet another transformation into ICICLE ( Incremental Computing In a Constructionist Learning Environment ). It is several things, but with an emphasis on introducing children to programming. It is quite impressive. Well, I think so anyway.
He has recently surveyed a sample of children in a local primary school to gauge the extent of their understanding of programming, with fascinating results.
If you're interested in this work, you should get in touch with Robert directly.
More exploits of Robert ?
I didn't share Natalie's supervision with George Blanchard of the Mechanical Engineering Department, but that was the original intention. For reasons which doubtless exist but remain unclear to me, George was not included in Natalie's enrolment, which happened while I was on leave in 1996.
Natalie continued the work she began in her earlier project work on programming for control systems using PDL. She was able to expand her implementation of PDL sufficiently to assess its viability as an industrial control systems programming method.
More PDL ?
More about Natalie ?
Some further information :
Ant's project was to develop various Internet and telephone based text-to-speech ( TTS ) and speech-to-text ( STT ) interfaces to provide gateways between the Internet and the Public Telephone Network. The intention is to use "Commercial Off-The-Shelf" software as far as possible : voice modem control software is available to play and record phone calls and process touch-tone codes, TTS and STT engines are available to process text and speech and WWW based POP account checking facilities are widely available over the Internet. All that is required in this project is to bind all the pieces together into a single package accessible via the Internet and Telephone networks.
Nothing is ever quite as easy as it sounds, but Ant was able to overcome complications posed by the mediocre quality and abilities of free or cheap software to produce something that demonstrably worked - provided that the recipients were warned what to expect.
Some further information :
Peter worked on a part of the "document factory" described in my working note ( see below ). He aimed to write a system in Java which would demonstrate the feasibility ( or infeasibility ) of the proposed means of implementing the "document factory" which appears at the end of the working note under the heading STRUCTURES AND IMPLEMENTATION.
It worked.
Some further information :
Maher's project was to persuade the mechanical turtle to learn to push a thing in a straight line. We had hoped that this might just lead to an interesting example for Paul Qualtrough's learning technique.
Unfortunately, problems of various sorts meant that he didn't get very far, and the expected work was never completed.
Some further information :
Louis worked on a remote ( internet ) control interface for driving the electric train.
This fitted into an investigation suggested by Paul Qualtrough, in which the train would be driven by an "intelligent" system which collects the incoming instructions, and would try to make sense of them, satisfying as many as it can. ( I think that was the idea - Paul wrote something about it, but that was a long time ago. ) It never happened, because Paul's ideas changed.
Some further information :
Lyndon's original intention was to use neural network ( or similar ) methods to find structure in EEG traces. This work is connected ( a lot more loosely than I originally expected, but that's life ) with the adaptive interface stream.
He ended up exploring the behaviour of learning vector quantisation ( LVQ ) methods in classifying segments of sine waves, which turned out to be very interesting. It must be admitted, though, that the applicability of his results to the original EEG problem was limited.
Some further information :
More about the adaptive interface ?
Jason worked on neural network controllers for stick figures.
He used a dynamic model of unconnected objects ( the sticks ), caused to behave as though they were connected by the application of impulsive forces and torques applied for every time interval of a simulation. The impulses were computed by neural network controllers. He devised means to train the controllers to perform standard actions - straighten leg, bend arm, etc. - and to coordinate these actions so that the stick figures could balance or walk. The results were very impressive.
From my point of view, it fits in with Chris Anderson's work on automatic production of sign language, as Jason's success should make it possible to complete the animation.
More animated signing ?
Alan Creak,
2005 November.
Go to me;
Go
to Computer Science.