PDF version of this document

Aims

The aims of the project were formally specified in Form 2 (see Appendix cha:App C:Year-1-Progress). The objectives set for the first year were slightly altered since the time of writing of the Literature Report^10.1. While in a rough sense, all objectives set were eventually accomplished, more goals and intermediary experiments were proposed and later completed. It was known earlier in the year that aims listed in Form 2 cannot be necessarily the right way to go, but only a formal and inductive set of guidelines. This is clearly because of the dependency which one experiment has upon another. Also, development within the Structure and Function group required attention to different aspects of work, as soon as issues began to arise. With reference to aims listed in past documents, many of these were not at all definite. From a non-specific^10.2 perspective, it was expected that:

1. A full reproduction of past experiments should be trivial and possible extensions realised.
2. Development of existing code will commence to ultimately build genuine software.
3. Difficulties should be identified to avoid future impasse.
4. New practicable experiments should be agreed upon, performed and their results recorded.
5. Comparative figures will show the advancements of new methods.
6. Critical evaluation of existing work and proposition of new methods will hopefully emerge.

Dealing with each of the above in turn, all previous experiments, as implemented by Smith, can be performed within seconds. A software package which was constructed was in fact enabling such experiments to be analysed in more depth whenever one requires so (referring back to Section sec:Environment).

Future difficulties were identified amongst the Structure and Function group and the project supervisor. These were frequently recorded and experiments that may suffer from such difficulties were avoided.

The new experiments were shown in previous chapters and comparative figures were an integral part of these. Not only was a benchmark for commonly-used methods established, but also a comparison - analytic and numeric - was made . Results amongst the different available alternatives justified the use of a model-based objective function as well.

Much more information about progress (and in finer level of detail) can be found in the weekly progress report and the sources that these reference. Experiment were very usefully put on an HTML-based database^10.3 and can be used as a supplementary resource for this document^10.4.

Proposition of new methods followed the analysis of different mutations of the model-based objective functions. However, nothing excitingly different or unexpected was discovered. It was merely the enhancement and modification of the basic model-based objective function that made it more powerful.