My understanding is that the randomisation step before optimisation is the same as the default behaviour of randomizeCoords(); "The maximum table distance is calculated then points are randomized in a box with side length equal to maximum table distance multiplied by table_dist_factor" where table_dist_factor=2. I tried relaxing a map with different table_dist_factor and found that 2 was not good (both in terms of stress and how the map looks - I have some not very well coordinated points that were getting hemisphered). table_dist_factor=0.1 had the lowest stress. table_dist_factor=0.01 had the next lowest (0.2% higher) and the map looked more consistent.
My ideal behaviour would be allowing the user to set the table_dist_factor when running optimizeMap(), perhaps as part of RacOptimizer.options. I don't know if the default should change, or if it's particular to large maps/those with lots of missing titres/this map. I expect a default based on the average distance would be better, as the maximum is affected by outliers.
My understanding is that the randomisation step before optimisation is the same as the default behaviour of
randomizeCoords(); "The maximum table distance is calculated then points are randomized in a box with side length equal to maximum table distance multiplied bytable_dist_factor" wheretable_dist_factor=2. I tried relaxing a map with differenttable_dist_factorand found that 2 was not good (both in terms of stress and how the map looks - I have some not very well coordinated points that were getting hemisphered).table_dist_factor=0.1had the lowest stress.table_dist_factor=0.01had the next lowest (0.2% higher) and the map looked more consistent.My ideal behaviour would be allowing the user to set the
table_dist_factorwhen runningoptimizeMap(), perhaps as part ofRacOptimizer.options. I don't know if the default should change, or if it's particular to large maps/those with lots of missing titres/this map. I expect a default based on the average distance would be better, as the maximum is affected by outliers.