mahout-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Suneel Marthi <suneel_mar...@yahoo.com>
Subject Re: Deprecated or drafts only algorithms what is the reasoning?
Date Fri, 03 Jan 2014 21:33:56 GMT





On Friday, January 3, 2014 4:26 PM, Dmitriy Lyubimov <dlieu.7@gmail.com> wrote:
 
On Fri, Jan 3, 2014 at 1:15 PM, Suneel Marthi <suneel_marthi@yahoo.com>wrote:

> See inline. The code for the deprecated algos has already been purged from
> trunk, its only the Wiki that needs cleaning up.
>
>
>
>
> On Friday, January 3, 2014 4:03 PM, "info@eprice.gr" <info@eprice.gr>
> wrote:
>
> Please confirm if below list  is correct before removing them:
>
> Classification
> Deprecated or drafts only:
>         * Support Vector Machines(see MAHOUT-14, MAHOUT-232and MAHOUT-334
>
>         * >>>>  Never made it  into the codebase and should not be
on
> Wiki, please remove all references to this from Wiki
>         * Perceptron and Winnow(see MAHOUT-85)
>         * >>>>  This has been removed from codebase for 0.9 due to lack
of
> usage and all references should be removed from Wiki
>
>         * Neural Network (see MAHOUT-228)
>         * >>>>  NEver made it to the codebase and all references should
be
> removed from Wiki
>
>         * Restricted Boltzmann Machines(see MAHOUT-375)
>         * >>>> Never made it to codebase and all references should be
> removed from Wiki
>
>         * Online Passive Aggressive (see MAHOUT-702
>         * Boosting (see MAHOUT-716)
>         * >>>> Same reasons as the previous ones, all references need
to
> be removed from the Wiki
>
> Clustering
> Deprecated or drafts only:
>         * Hierarchical Clustering (MAHOUT-19, MAHOUT-843)
>
>         * Spectral Clustering (MAHOUT-363)
>         * >>>> This should be retained, Shannon will be updating the
Wiki
> for this later.
>
> Dimension reduction
> Deprecated or drafts only:
>         * Principal Components Analysis (PCA)
>

PCA should probably be rewritten to point to SSVD with PCA options (SSVD
solver includes end2end PCA functionality)

> If the Wiki for SSVD already covers SSVD with PCA options, may be it may make sense to
purge the older PCA only wiki page.


>         * >>>> We still have this and its widely used, retain this.
>
>         * Gaussian Discriminative Analysis (GDA)
>         * >>>> Retain this.
>
> Other
> Deprecated or drafts only:
>         * Pattern mining: Parallel FP Growth Algorithm(Also known as
> Frequent Itemset mining)
>         * >>> Code for this has been removed from 0.9, all references to
> this should be removed from Wiki
>

On not directly related plane, could somebody do me a favor and tell why FP
Growth was removed? Efficiency issues? Lack of maintenance? thanks.



>
>
>
>
> From:Suneel Marthi [mailto:suneel_marthi@yahoo.com]
> Sent: Friday, January 03, 2014 10:48 PM
> To: dev@mahout.apache.org; info@eprice.gr
> Subject: Re: Deprecated or drafts only algorithms what is the reasoning?
>
> It could be because:-
>
> a) they have been replaced by better performant alternatives
> b) lack of usage
> c) lack of support
>
> Please delete from wiki all algorithms that have been marked deprecated
> (the code for most of them has already been removed from trunk).
>
>
> On Friday, January 3, 2014 3:31 PM, "info@eprice.gr" <info@eprice.gr>
> wrote:
> Hi all,
>
>
>
> I'm going through the algorithms and I noticed some are "Deprecated or
> drafts only" . What is the reason?
>
>
>
> Is it because they are less efficient compared to the supported ones or is
> it due to lack of resources working on them to be supported with Mahout?
>
>
>
> Thanks
>
> Sotiris
>
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message