Opened 9 years ago

Last modified 2 years ago

#1386 new enhancement

Mutation of constant factors should not simply add a normally distributed value but a value sampled from a mixture of Gaussians

Reported by: gkronber Owned by: gkronber
Priority: low Milestone: HeuristicLab 4.0
Component: Problems.DataAnalysis Version: 3.3.2
Keywords: Cc: maffenze

Description

Now mutation of constants and variable weights is:

f = f + x
x ~ N(0,sigma),

where sigma is a parameter supplied by the user.

This should be changed as suggested by maffenze to the following more complex adaption routine that is arguably more robust.

f = f + y
y ~ N(0,s)
s ~ U(0, s_max),

where s_max is a parameter supplied by the used. For variables s_max should automatically set to:

 s_max = s_user * Var(x_i),

Where s_user is a user supplied parameter and Var(x_i) is the variance of variable x_i over the observations in the training set.

Change History (3)

comment:1 Changed 9 years ago by swagner

  • Milestone changed from HeuristicLab 3.3.3 to HeuristicLab x.x.x

comment:2 Changed 7 years ago by gkronber

  • Priority changed from medium to low
  • Summary changed from Mutation of constant factors should not simply add a random normally distributed variable but a random variable that is a mixture of normally distributed variables to Mutation of constant factors should not simply add a normally distributed value but a value sampled from a mixture of Gaussians

comment:3 Changed 2 years ago by gkronber

  • Milestone changed from HeuristicLab 3.3.x Backlog to HeuristicLab 4.0

I suggest we use the switch to the new version (allowing behavior changes) to implement this change.

An quick experiment should be made first to check that algorithm performance doesn't deteriorate.

Note: See TracTickets for help on using tickets.