mahout-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dmitriy Lyubimov <dlie...@gmail.com>
Subject Re: EigenDecomposition
Date Wed, 10 Jul 2013 21:13:59 GMT
meant, after code freeze is over.


On Wed, Jul 10, 2013 at 2:13 PM, Dmitriy Lyubimov <dlieu.7@gmail.com> wrote:

> fixed as part of MAHOUT-1281 patch now. I will push after code freeze.
>
>
> On Wed, Jul 10, 2013 at 2:06 PM, Ted Dunning <ted.dunning@gmail.com>wrote:
>
>> Please file.  Looks completely innocuous and it is good to be standard.
>>
>>
>> On Wed, Jul 10, 2013 at 12:59 PM, Dmitriy Lyubimov <dlieu.7@gmail.com
>> >wrote:
>>
>> > Looks like Lanczos is having the same problem and need to undo some
>> > workarounds :
>> >
>> >     EigenDecomposition decomp = new EigenDecomposition(triDiag);
>> >
>> >     Matrix eigenVects = decomp.getV();
>> >     Vector eigenVals = decomp.getRealEigenvalues();
>> >     endTime(TimingSection.TRIDIAG_DECOMP);
>> >     startTime(TimingSection.FINAL_EIGEN_CREATE);
>> >     for (int row = 0; row < i; row++) {
>> >       Vector realEigen = null;
>> >       // the eigenvectors live as columns of V, in reverse order.  Weird
>> > but true.
>> >       Vector ejCol = eigenVects.viewColumn(i - row - 1);
>> >       int size = Math.min(ejCol.size(), state.getBasisSize());
>> >
>> >
>> >
>> > On Wed, Jul 10, 2013 at 12:53 PM, Dmitriy Lyubimov <dlieu.7@gmail.com
>> > >wrote:
>> >
>> > > changing line 329 of EigenDecomposition.java from
>> > >
>> > >         if (d.getQuick(j) < p) {
>> > >
>> > > to
>> > >         if (d.getQuick(j) > p) {
>> > >
>> > >
>> > > makes my MAHOUT-1281 patch work.
>> > >
>> > > should i keep the change? (question for Ted, i guess)
>> > >
>> > > thanks.
>> > > -D
>> > >
>> > >
>> > >
>> > >
>> > > On Wed, Jul 10, 2013 at 11:59 AM, Dmitriy Lyubimov <dlieu.7@gmail.com
>> > >wrote:
>> > >
>> > >> It looks like values out of our ported EigenDecomposition are coming
>> out
>> > >> sorted in inverse order.
>> > >>
>> > >> Shouldn't it be the other way around?
>> > >>
>> > >
>> > >
>> >
>>
>
>

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