You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Jason Thomas <ka...@gmail.com> on 2012/08/11 15:37:00 UTC
genetic algorithms / watchmaker removal
I noticed Watchmaker support was removed. Could someone tell me if this
was because genetic algorithms in general aren't in line with Mahout goals
or was there a problem specifically with the Watchmaker
integration/implementation? I need scalable GA and I'm trying to figure
out how to proceed.
Thanks,
Jason
Re: genetic algorithms / watchmaker removal
Posted by Ted Dunning <te...@gmail.com>.
Holler if you need help with the API. There is some templating that is
done to allow generality of the optimization. That might make proper use
somewhat obscure.
On Sun, Aug 12, 2012 at 4:14 PM, Jason Thomas <ka...@gmail.com>wrote:
> Thanks for the help Ted, I will check it out.
>
> On Sun, Aug 12, 2012 at 1:36 AM, Ted Dunning <te...@gmail.com>
> wrote:
>
> > That sounds like a continuous optimization problem.
> >
> > Look at the org.apache.mahout.ep.EvolutionaryProcess
> >
> > It is an implementation of recorded step meta-mutation and does quite
> well
> > on many problems.
> >
> > See http://arxiv.org/abs/0803.3838 for details on the algorithm.
> >
> > On Sat, Aug 11, 2012 at 8:24 PM, Jason Thomas <katsoftwarellc@gmail.com
> > >wrote:
> >
> > > It's a search/optimization problem. I have an existing complex model
> > > (multiple neural nets) and I need to find the right combination of
> inputs
> > > to achieve certain target outputs. It seems like GA would be the
> easiest
> > > approach for this.
> > >
> > > I'll take a look at the old Watchmaker code and maybe try to improve on
> > it.
> > > Thanks for the help.
> > >
> > > -Jason
> > >
> > >
> > > On Sat, Aug 11, 2012 at 6:20 PM, Ted Dunning <te...@gmail.com>
> > > wrote:
> > >
> > > > The Watchmaker implementation was not very scalable and there was no
> > > > perceptible user demand for it. There was also nobody who was
> > > maintaining
> > > > it.
> > > >
> > > > So we nuked it.
> > > >
> > > > There is still a limited evolutionary algorithm that is part of the
> > > > AdaptiveLogisticRegression. It is likely to be pretty good on
> problems
> > > > that involve the optimization of functions of continuous variables.
> > > >
> > > > What sort of GA do you need?
> > > >
> > > > On Sat, Aug 11, 2012 at 6:37 AM, Jason Thomas <
> > katsoftwarellc@gmail.com
> > > > >wrote:
> > > >
> > > > > I noticed Watchmaker support was removed. Could someone tell me if
> > > this
> > > > > was because genetic algorithms in general aren't in line with
> Mahout
> > > > goals
> > > > > or was there a problem specifically with the Watchmaker
> > > > > integration/implementation? I need scalable GA and I'm trying to
> > > figure
> > > > > out how to proceed.
> > > > >
> > > > > Thanks,
> > > > > Jason
> > > > >
> > > >
> > >
> >
>
Re: genetic algorithms / watchmaker removal
Posted by Jason Thomas <ka...@gmail.com>.
Thanks for the help Ted, I will check it out.
On Sun, Aug 12, 2012 at 1:36 AM, Ted Dunning <te...@gmail.com> wrote:
> That sounds like a continuous optimization problem.
>
> Look at the org.apache.mahout.ep.EvolutionaryProcess
>
> It is an implementation of recorded step meta-mutation and does quite well
> on many problems.
>
> See http://arxiv.org/abs/0803.3838 for details on the algorithm.
>
> On Sat, Aug 11, 2012 at 8:24 PM, Jason Thomas <katsoftwarellc@gmail.com
> >wrote:
>
> > It's a search/optimization problem. I have an existing complex model
> > (multiple neural nets) and I need to find the right combination of inputs
> > to achieve certain target outputs. It seems like GA would be the easiest
> > approach for this.
> >
> > I'll take a look at the old Watchmaker code and maybe try to improve on
> it.
> > Thanks for the help.
> >
> > -Jason
> >
> >
> > On Sat, Aug 11, 2012 at 6:20 PM, Ted Dunning <te...@gmail.com>
> > wrote:
> >
> > > The Watchmaker implementation was not very scalable and there was no
> > > perceptible user demand for it. There was also nobody who was
> > maintaining
> > > it.
> > >
> > > So we nuked it.
> > >
> > > There is still a limited evolutionary algorithm that is part of the
> > > AdaptiveLogisticRegression. It is likely to be pretty good on problems
> > > that involve the optimization of functions of continuous variables.
> > >
> > > What sort of GA do you need?
> > >
> > > On Sat, Aug 11, 2012 at 6:37 AM, Jason Thomas <
> katsoftwarellc@gmail.com
> > > >wrote:
> > >
> > > > I noticed Watchmaker support was removed. Could someone tell me if
> > this
> > > > was because genetic algorithms in general aren't in line with Mahout
> > > goals
> > > > or was there a problem specifically with the Watchmaker
> > > > integration/implementation? I need scalable GA and I'm trying to
> > figure
> > > > out how to proceed.
> > > >
> > > > Thanks,
> > > > Jason
> > > >
> > >
> >
>
Re: genetic algorithms / watchmaker removal
Posted by Ted Dunning <te...@gmail.com>.
That sounds like a continuous optimization problem.
Look at the org.apache.mahout.ep.EvolutionaryProcess
It is an implementation of recorded step meta-mutation and does quite well
on many problems.
See http://arxiv.org/abs/0803.3838 for details on the algorithm.
On Sat, Aug 11, 2012 at 8:24 PM, Jason Thomas <ka...@gmail.com>wrote:
> It's a search/optimization problem. I have an existing complex model
> (multiple neural nets) and I need to find the right combination of inputs
> to achieve certain target outputs. It seems like GA would be the easiest
> approach for this.
>
> I'll take a look at the old Watchmaker code and maybe try to improve on it.
> Thanks for the help.
>
> -Jason
>
>
> On Sat, Aug 11, 2012 at 6:20 PM, Ted Dunning <te...@gmail.com>
> wrote:
>
> > The Watchmaker implementation was not very scalable and there was no
> > perceptible user demand for it. There was also nobody who was
> maintaining
> > it.
> >
> > So we nuked it.
> >
> > There is still a limited evolutionary algorithm that is part of the
> > AdaptiveLogisticRegression. It is likely to be pretty good on problems
> > that involve the optimization of functions of continuous variables.
> >
> > What sort of GA do you need?
> >
> > On Sat, Aug 11, 2012 at 6:37 AM, Jason Thomas <katsoftwarellc@gmail.com
> > >wrote:
> >
> > > I noticed Watchmaker support was removed. Could someone tell me if
> this
> > > was because genetic algorithms in general aren't in line with Mahout
> > goals
> > > or was there a problem specifically with the Watchmaker
> > > integration/implementation? I need scalable GA and I'm trying to
> figure
> > > out how to proceed.
> > >
> > > Thanks,
> > > Jason
> > >
> >
>
Re: genetic algorithms / watchmaker removal
Posted by Jason Thomas <ka...@gmail.com>.
It's a search/optimization problem. I have an existing complex model
(multiple neural nets) and I need to find the right combination of inputs
to achieve certain target outputs. It seems like GA would be the easiest
approach for this.
I'll take a look at the old Watchmaker code and maybe try to improve on it.
Thanks for the help.
-Jason
On Sat, Aug 11, 2012 at 6:20 PM, Ted Dunning <te...@gmail.com> wrote:
> The Watchmaker implementation was not very scalable and there was no
> perceptible user demand for it. There was also nobody who was maintaining
> it.
>
> So we nuked it.
>
> There is still a limited evolutionary algorithm that is part of the
> AdaptiveLogisticRegression. It is likely to be pretty good on problems
> that involve the optimization of functions of continuous variables.
>
> What sort of GA do you need?
>
> On Sat, Aug 11, 2012 at 6:37 AM, Jason Thomas <katsoftwarellc@gmail.com
> >wrote:
>
> > I noticed Watchmaker support was removed. Could someone tell me if this
> > was because genetic algorithms in general aren't in line with Mahout
> goals
> > or was there a problem specifically with the Watchmaker
> > integration/implementation? I need scalable GA and I'm trying to figure
> > out how to proceed.
> >
> > Thanks,
> > Jason
> >
>
Re: genetic algorithms / watchmaker removal
Posted by Ted Dunning <te...@gmail.com>.
The Watchmaker implementation was not very scalable and there was no
perceptible user demand for it. There was also nobody who was maintaining
it.
So we nuked it.
There is still a limited evolutionary algorithm that is part of the
AdaptiveLogisticRegression. It is likely to be pretty good on problems
that involve the optimization of functions of continuous variables.
What sort of GA do you need?
On Sat, Aug 11, 2012 at 6:37 AM, Jason Thomas <ka...@gmail.com>wrote:
> I noticed Watchmaker support was removed. Could someone tell me if this
> was because genetic algorithms in general aren't in line with Mahout goals
> or was there a problem specifically with the Watchmaker
> integration/implementation? I need scalable GA and I'm trying to figure
> out how to proceed.
>
> Thanks,
> Jason
>