Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: average change in the spread of Pareto solutions less than options
Posted:
Aug 19, 2013 3:17 PM


On 8/17/2013 4:42 AM, guymcheers wrote: > Who can tell me this following optimization statement is successful or > fail, when I use the gamultiobj toolbox? > > "Optimization terminated: average change in the spread of Pareto > solutions less than options.TolFun." > > If it means fail, what is the reason? > > If it means successful, what does it mean? > > "TolFun" means what? > > > Many thanks:) > > > guymcheers!
This is a good exit message, but you have found some deficiencies in the documentation that I aim to address in the future. The documentation does not describe what "spread of Pareto solutions" is. This is a notion in Kalyanmoy Deb's book "MultiObjective Optimization using Evolutionary Algorithms." Basically, the idea is that the Pareto front is not changing much toward the end of the optimization, so the solver stops.
TolFun is a tolerance described in the documentation http://www.mathworks.com/help/optim/ug/tolerancesandstoppingcriteria.html The TolFun tolerance in this case is measured against the change of the spread of the solutions.
Sorry for the deficiencies in the documentation, I'll try to improve it.
Alan Weiss MATLAB mathematical toolbox documentation



