[R] Packages for Learning Algorithm Independent Branch and Bound for Feature Selection
anbyrley at buffalo.edu
Sat Jul 1 22:42:57 CEST 2017
See, I have built my own genetic algorithm already and tested it on this
problem. I have a solution, but due to the heuristic nature of GA, I cannot
guarantee that it is the optimal subset.
If I was simply doing this for a company project, you are spot on with the
type of algorithm I would use, but I am doing this for a scientific paper.
I need to be able to find the optimal subset over my dataset, and I know
branch and bound will find it without resorting to exhaustive search. If I
can't claim that my subset is optimal, it is going to open my paper up to
serious enough criticism that it will get rejected, regardless of whether
my new method outperforms the state-of-the-art or not. (And regardless if
my dataset is representative enough to make such performance claims)
I will take a look at that page, thanks! Hopefully there is an R
implementation of generic B&B as I described out there somewhere...
Department of Electrical Engineering
235 Davis Hall
2017-07-01 3:53 GMT-04:00 Enrico Schumann <es at enricoschumann.net>:
> On Thu, 29 Jun 2017, Alex Byrley writes:
> > I am looking for packages that can run a branch-and-bound algorithm to
> > maximize a distance measure (such as Bhattacharyya or Mahalanobis) on a
> > of features.
> > I would like this to be learning algorithm independent, so that the
> > just looks at the features, and selects the subset of a user-defined size
> > that maximizes a distance criteria such as those stated above.
> > Can anyone give some suggestions?
> > Alex Byrley
> > Graduate Student
> > Department of Electrical Engineering
> > 235 Davis Hall
> > (716) 341-1802
> It seems you are looking for a generic optimisation
> algorithm; so perhaps start at the task view:
> What you describe is a combinatorial problem: select k
> from N features, with k (much) smaller than N. So I'd
> suggest to also look at heuristic algorithms that can
> deal with such problems (e.g. genetic algorithms).
> Enrico Schumann
> Lucerne, Switzerland
[[alternative HTML version deleted]]
More information about the R-help