PRSD Studio

classifier output as support function

2010-03-10T14:12:47Z

Hello,
I am relatively new to prtools -learning in progress:)

I’m looking for a way to achieve from classifier (mapping) not direct class label but a support function value in case, for instance, neural nets - or probabilities that given object is of class X. Generally I want to achieve soft value from given classifier. Is it easly possible or requires a lot of coding?
Is it correct assumption that after operation dataset * mapping I should get a new dataset where
resulting nlab array should contain soft labels set by mapping?

Kind regards,
Tomasz Wilk

measures in sdcrossval

2010-03-05T16:14:51Z

Hi,

I would like to know the list of ‘measures’ that can be used in sdcrossval. In particular I would like to find a measure equivalent to:

res=sdcrossval(sdknn*sddecide, data, 'measures',{'class-errors'})*getpriors(data)'

Thanks

Mathieu

how to set parameter using treec() when wanting to applying C4.5 in boosting?

2010-03-09T05:05:13Z

Hi, I want to use C4.5 algorithm in adaboost for comparison with the well known published works related to boosing papers for the fact that C4.5 is usually used in previous works,
How do I set the parameter for function treec in Prtools in order to perform the classic C4.5 algorithm?

thanks

Can’t change the cross-validation strategy in sdcrossval

2010-03-08T17:05:23Z

Hi,

In PRSD Studio 2.0.5, I get the following error when I tried the example in the documentation about the different cross-validation strategies:

>> load fruit; pd= sdgauss*sddecide; [r2,e2]=sdcrossval(pd,a,'folds',2,'method','random');
??? Undefined function or method 'getstate' for input arguments of type 'sdppl'.

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\private\sde_randomization.p>sde_randomization at 19

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdcrossval.p>sdcrossval at 120

Thank you for your help.

Mathieu

Use custom feature selection algorithm in a pipeline

2010-03-05T16:08:31Z

Hi,

I created a custom feature selection algorithm with sdalg. Is it possible to use it in a pipeline so that I can do the following thing:


p = sdcrossval((myalgo*sdlinear)*sddecide,data)

I tried to use sdp_fsel but I got the following error when trying it:

>> p=sdp_fsel(14, [1 2 4 6])
??? Input argument "lab" is undefined.

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdp_fsel.p>sdp_fsel at 24

Thanks

Mathieu

sdtree error

2010-03-05T15:41:53Z

Hi,

I use PRSD Studio 2.0.5 and I have the following error when I use the sdtree classifier on the adult dataset from UCI repository (in attachment): (Note: the problem seems to happen every time my dataset has more than 19 rows)

>> sdtree(data)
.??? Undefined function or method 'getcodes' for input arguments of type 'double'.

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\@sdlab\subset.p>subset at 23

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\@sddata\private\subset_ind.p>subset_ind at 17

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\@sddata\subsref.p>subsref at 35

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdtree.p>tree_getcrit at 288

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdtree.p>tree_featthr at 281

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdtree.p>tree_processallfeat at 250

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdtree.p>tree_train at 140

Error in ==> C:\Program Files\MATLAB\R2009b\toolbox\prsd_toolbox\sdtree.p>sdtree at 49

bpxnc - how this works

2010-02-28T21:32:27Z

Hallo,
I am trying to understand the way the bpxnc works. I read documentation and right now I am trying to get closer the the PRTools code.
Could You please send me the picture (jpg) how networks structure looks like when I type bpxnc([],[10 10], nr_of_epoche)? I need just the draft picture in which I could find the information where are weights and how the input and output of such network looks like for the ‘soft’ version.

[10 10] means number of units in each hidden layer

Please help me with this problem

votec

2010-03-06T16:40:42Z

hello prtools team
I am trying to combine diffrent base classifiers using majority voting ‘votec’. The classifiers are: ldc,knnc, svc and treec. It work with all the combinig rule prodc, minc, medianc, maxc, meanc but with votec i have this error message:

index exceeds matrix dimensions.

Error in ==> C:\MATLAB6p5\work\prtools\@dataset\subsref.m
On line 78 ==>

Error in ==> C:\MATLAB6p5\work\prtools\fixedcc.m
On line 143 ==> labels = labeld(a(:,J));

Error in ==> C:\MATLAB6p5\work\prtools\fixedcc.m
On line 177 ==> v = feval(mfilename,a*w.data{1},[],type,name);

Error in ==> C:\MATLAB6p5\work\prtools\map.m
On line 179 ==> d = feval(mapp,a,b);

Error in ==> C:\MATLAB6p5\work\prtools\@mapping\mtimes.m
On line 15 ==>

Error in ==> C:\MATLAB6p5\work\prtools\testc.m
On line 289 ==> [out1,out2]= feval(mfilename,a*w,type,label);[/size]

thanks in advance

Turn off messages in sdcrossval

2010-03-05T17:23:00Z

Hi,

Is there a way to turn off messages like: 10 folds: [1: ] [2: ] ... in sdcrossval?

Thanks

Mathieu

sda_featex is missing

2010-03-05T15:38:08Z

Hi,

It seems that the sda_featex function, mentioned in the knowledge base, is missing in PRSD Studio 2.0.5.

Mathieu