This page exists to give status reports and documentation of the Fall 2009 comparison between Iterated F-RACE and ParamILS on various problem domains. These domains and their scenarios are as follows:

Solution-Quality Optimisation

Ant-Colony Optimisation for the TSP

7 Parameter Case

• Tuned for both 5 second runs and 20 second runs.
• Tuned for 768, 1728 and 3888 function evaluations (algorithm runs).

Parameter Name	Domain
alpha	{0.01, 0.5, 1.0, 2.0, 5.0}
beta	{0.0, 0.5, 1.0, 2.0, 5.0, 10.0}
rho	{0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 0.8, 1.0}
ants	{1, 5, 10, 50, 100}
nnants	{5, 10, 20, 40}
q0	{0.0, 0.02, 0.1, 0.2, 0.5, 0.8, 0.9, 0.98, 1.0}
time	{0.01, 0.1, 0.5, 1.0, 2.0, 5.0}

12 Parameter Case

• Tuned for both 5 second runs and 20 second runs.
• Tuned for 1500, 3000 and 6000 function evaluations (algorithm runs).

Parameter Name	Domain	Conditions
alpha	{0.01, 0.5, 1.0, 2.0, 5.0}
beta	{0.0, 0.5, 1.0, 2.0, 5.0, 10.0}
rho	{0.0001, 0.02, 0.05, 0.1, 0.2, 0.5, 0.8, 1.0}
ants	{1, 5, 10, 50, 100}
nnants	{3, 5, 10, 20, 40, 100, 200, 500}
q0	{0.0, 0.02, 0.1, 0.2, 0.5, 0.8, 0.9, 0.98, 1.0}	Active when mode={acs}
nnls	{3, 5, 10, 20, 40, 100, 200, 500}	Active when localsearch={1,2,3}
rasrank	{1, 2, 4, 6, 8, 10}	Active when mode={ras}
localsearch	{0, 1, 2, 3}
dlb	{0,1}	Active when localsearch={1,2,3}
mode	{mmas, acs, ras, eas, as}
elitistants	{1, 5, 10, 50, 100}	Active when mode={eas}

Algorithms for the Probabilistic TSP

Runtime Optimisation

Post-Enrollment Course Timetabling

SATenstein

CPLEX

-- ChrisFawcett - 09 Sep 2009