20/1/2005 - Extended - Appended Progress

___________________________________________
Mass Experiments

-Shut-down of computers during the vacation.
-Experiments run locally, in ISBE.
-January 4th: A new working build of the VXL code on the CS domain in.


___________________________________________
Optimisation

-Simplex - good results obtained quickly. Downhill search susceptible to local minima.
-Powell has been looked at as well and it is similar to simplex in terms of time and quality.
-See pictures of comparisons in 3-D.
-Intent was to discover better optimisation and Powell is believed to be a good match.
-The conclusion is that Powell should be used given more time that is 'allocated' for optimisation.


___________________________________________
Models and Registration

-Implementation of model determinant as OF has begun.
-Code progressively put on the CVS repository.
-APM module familiarity and understanding still needed.


___________________________________________
3-D Groupwise NRR - Histogram-Based Methods

-MI attempted in 3-D but results were poor, as before.
-Res_info with the new code in 3-D
-Failed to get any registration at all. Many bin filters and numbers of bins attempted.