mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel Zohar <disso...@gmail.com>
Subject Re: Mahout performance issues
Date Fri, 02 Dec 2011 18:43:29 GMT
Sean, can you tell me which files have you committed the changes to? Thanks

On Fri, Dec 2, 2011 at 8:22 PM, Sean Owen <srowen@gmail.com> wrote:

> Yes, and there is no assumption here that the item is vastly popular. The
> proposal was to drop any user-item interaction where the item was the only
> one rated by the user, whether the item is popular or not. There's a
> non-zero, and probably non-trivial, affect on correctness as a result.
>
> On Fri, Dec 2, 2011 at 6:18 PM, Ted Dunning <ted.dunning@gmail.com> wrote:
>
> > On Fri, Dec 2, 2011 at 10:14 AM, Sean Owen <srowen@gmail.com> wrote:
> >
> > > I agree, though that's something slightly different to the question at
> > hand
> > > here. If just about every user viewed X, you could probably forget X,
> > yes.
> > >
> > > But if some user happened to only view one item, Y, can you drop that?
> It
> > > affects correctness. I argue that it doesn't really affect the
> bottleneck
> > > in question here, which is not quite the point you are getting at.
> > >
> >
> > No.  Dropping that does not affect correctness if that item is highly
> > popular.
> >
> > If that item is not vastly popular, it does provide some information and
> > would not be affected by my suggestions.
> >
> >
> > >
> > > On Fri, Dec 2, 2011 at 6:07 PM, Ted Dunning <ted.dunning@gmail.com>
> > wrote:
> > >
> > > > Since touching them adds nothing but cost, then not touching them is
> > > > better.  Kill the item!
> > > >
> > > > In practical terms, we had this problem at Veoh.  Everybody got the
> > same
> > > > intro video.  It provided no information.  Likewise at Musicmatch,
> > > > everybody got the same startup noise during the splash screen.  It
> > added
> > > no
> > > > information.  Both of these cases would kill performance in lots of
> > > > recommendation engines because a vast number of users would get
> sucked
> > > into
> > > > computations where it made no difference at all.
> > > >
> > > > Better to kill these items.
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message