Experimental Results

Complementing the case studies, we performed experiments to better understand the power of randomization and our cooling schedules as referenced in Table 4.2.

Table 4.2: Key to Methods

Abbreviation	Description
sa0	algorithmic
sa0/m	algorithmic, + min step
sa1	table-driven
sa1/m	table-driven, + min step
sa2	attempt-based adaptive
sa2/m	attempt-based adaptive, + min step
sa3	accept-based adaptive
sa3/m	accept-based adaptive, + min step
sa4	reheating accept-based adaptive
sa4/m	reheating accept-based adaptive, + min step
sigmet1	Metropolis and table-driven schedule (half laterals)
sigmet1/m	Metropolis and table-driven schedule, + min step
sigmet4	Metropolis and reheating schedule (half laterals)
sigmet4/m	Metropolis and reheating schedule, + min step
met1	Metropolis and table-driven schedule (full laterals)
met1/m	Metropolis and table-driven schedule, + min step
met4	Metropolis and reheating schedule (full laterals)
met4/m	Metropolis and reheating schedule, + min step
rdhc	random descent hillclimbing (no laterals)
rdhc/m	random descent hillclimbing, + min step
rldhc	random lateral-descent hillclimbing (half laterals)
rldhc/m	random lateral-descent hillclimbing, + min step
rldhc2	random lateral-descent hillclimbing (full laterals)
rldhc2/m	random lateral-descent hillclimbing, + min step
mr	minimize using interated next-descent hillclimbing
	(random start in arbitrary order)
m	minimize using interated next-descent hillclimbing
	(in priority order)
r	randomized

download thesis