Re: [Clus-devel] Clus Usage

[Bernard Z. <ber...@ij...>]

Dear Mohamed,

sorry for late reply. I will try to answer what I know and I would like 
to ask others to correct me, because I'm only familiar with parts of 
Clus related to rule induction.

> * how  attributes weights are used in the algorithm?

The attribute's weights are meant to put more or less emphasis to 
specific attributes during the learning phase. How exactly this is done 
depends on the algorithm and the heuristic used. In case of rule 
learning, the weights are used for clustering attributes (these are the 
attributes that are considered during evaluation of clusters, in most 
cases the clustering attributes are equal to the target attributes). The 
overall heuristic value consists of components of separate attributes 
that are weighted by these weights. At the moment the weights are not 
used for descriptive attributes, e.g., to force certain attributes into 
condition of rules. As far as I know this is also true for trees, but 
not necessarily for all heuristics (best to have a look at the code of 
the specific heuristic to be sure).

> * are xval different from "-fold" ? the definition is almost the same
> in the guide

-xval option means that the entire cross-validation is performed (1x on 
the whole training data and 10x on each of the folds). -fold, on the 
other hand, is meant to only run one fold of the cross-validation, i.e., 
on the fold N. This is useful, e.g., if you want to send each fold to 
another computer to do the cross-validation in parallel.

> how to view the model file that gets created when the prediction
> file get created ?

This I don' know (never used it myself), but I think that the .model 
file is in a binary file and can only be read programmatically to load a 
model. Best to inspect the code if you want to use this.

I hope this at least partially answers your questions.

Regards, Bernard


On 14.3.14 23:50, Mohamed Ayed wrote:
> Hi All,
>
> how to view the model file that gets created when the prediction file
> get created ?
>
> Thanks
> Mohamed
>
>
> On Thu, Mar 13, 2014 at 4:45 PM, Mohamed Ayed <mh...@gm...
> <mailto:mh...@gm...>> wrote:
>
>     Hi Bernard,
>
>     Thanks for the reply. I will be using the tool for drug discovery
>     purposes.
>
>
>     I definitely have some questions...here are some for a start:
>
>
>       * how  attributes weights are used in the algorithm?
>       * are xval different from "-fold" ? the definition is almost the
>         same in the guide
>
>
>     Thanks
>     Mohamed
>
>
>
>     On Wed, Mar 12, 2014 at 9:41 AM, Bernard Zenko <ber...@ij...
>     <mailto:ber...@ij...>> wrote:
>
>         Dear Mohamed,
>
>         unfortunately, what you say is true :) At the moment there is no
>         better
>         documentation than this. I suggest that if you have a question
>         regarding
>         a specific option or setting, you post it here and hopefully the
>         author
>         that introduced it will explain it.
>
>         Best, Bernard
>
>
>         On 10.3.14 9:01, Mohamed Ayed wrote:
>          > Hi All,
>          >
>          > I have been reading the document about how to use the
>         different settings
>          > and variables in the settings and .arff  files. Sometimes the
>          > description for each field does not give a clear
>         understanding of its
>          > usage and there are few examples on the usage of each field
>         or setting.
>          > Is there a source for more explanations and possibly more
>         different
>          > usage and tuning examples ?. I have also looked at the data
>         directory
>          > that comes with the package. They are good examples but lack the
>          > description of the indicated behaviors associated for the
>         indicated
>          > settings or attributes and that is I what I am looking for.
>          >
>          > Thanks
>          > Mohamed
>          >
>          >
>          >
>         ------------------------------------------------------------------------------
>          > Learn Graph Databases - Download FREE O'Reilly Book
>          > "Graph Databases" is the definitive new guide to graph
>         databases and their
>          > applications. Written by three acclaimed leaders in the field,
>          > this first edition is now available. Download your free book
>         today!
>          > http://p.sf.net/sfu/13534_NeoTech
>          >
>          >
>          >
>          > _______________________________________________
>          > Clus-devel mailing list
>          > Clu...@li...
>         <mailto:Clu...@li...>
>          > https://lists.sourceforge.net/lists/listinfo/clus-devel
>          >
>
>         ------------------------------------------------------------------------------
>         Learn Graph Databases - Download FREE O'Reilly Book
>         "Graph Databases" is the definitive new guide to graph databases
>         and their
>         applications. Written by three acclaimed leaders in the field,
>         this first edition is now available. Download your free book today!
>         http://p.sf.net/sfu/13534_NeoTech
>         _______________________________________________
>         Clus-devel mailing list
>         Clu...@li...
>         <mailto:Clu...@li...>
>         https://lists.sourceforge.net/lists/listinfo/clus-devel
>
>
>
>
>     --
>     Thanks
>     Mohamed
>
>
>
>
> --
> Thanks
> Mohamed