Determining Best Search Parameters
From TNT
For many data sets, the time to find best scores during searches may depend dramatically on the aggresivenes and parameters used during the search. For example, using very superficial searches for difficult data sets will keep grinding for a very long time before the best score is actually found; using more aggressive parameters will find the best score sooner. The parameters that can be used to change the degree of agressiveness are size and number of sectors to run under random sectorial searches, number of initial replications to submit to fusing, and rounds of tree-drifting or ratchet.
A careful consideration of how the data set behaves to changes in parameters is the best way to proceed, but if you have no idea of how a data set behaves, or don't know how to set the parameters for a search, you may trust to the program the choice of parameters. This is done with the set initial level option of the New Technology dialog box (or the level option of the xmult command). If entered as a command, the level must be a number between 0 and 10 (0 is very superficial; 10 is extremely exhaustive); if entered in the dialog box for search level, it must be a number between 0 and 100 (again, 100 is the heaviest search). If using a driven search (i.e. more than a single hit of the xmult command), you may have the program check whether best score is being found easily or not (and then decrease or increase the search level accordingly). This is set with the chklevel option of the xmult command. Again, letting the program try to determine the best parameters during the search is a poor substitute for finding the best parameters and letting them fixed during the search, and is intended for cases where you have no time/experience to determine best parameters by yourself.
Path: Analyze/New Technology Search
