PRSD Studio

Dataset addition of a user selected grouping variable.

2008-12-22T16:56:50Z

Hi again PRteam, I found useful to have a grouping variable in my feature matrix to separate feature vectors extracted from different experiments, in my case different pacients. The best way I found to implement this feature was with the USER field, and the update of this variable was done in my application, but I suppose that could be implemented like a dataset property, and some methods like SUBREF, etc could update this variable automatically. I used this variable to perform cross-validation in a leave-one-pacient-out fashion, and different training-testing aggrupations. Maybe there is a best way to do this with the existing features ... what do you think?

Mariano.

Problem using mixture of gaussians (gaussm)

2008-12-19T18:03:41Z

Hi everybody. I am working with mixture of gaussians models for multi-class classification and found that these models always maps from ‘k’ to 1 being k the number of features in the dataset. This can be seen in the example “prex_mcplot.m” comparing the trained mapping:

w = qdc(A)
Bayes-Normal-2, 2 to 4 trained mapping --> normal_map

for example with:

w2 = gaussm(A,3)
Mixture of Gaussians, 2 to 1 trained mapping --> sequential

And if I try:

confmat(A*w)

True | Estimated Labels
Labels | 1 2 3 4 | Totals
--------|----------------------------|-------
1 | 19 1 0 0 | 20
2 | 0 20 0 0 | 20
3 | 0 0 20 0 | 20
4 | 0 0 0 20 | 20
--------|----------------------------|-------
Totals | 19 21 20 20 | 80

while if I try:

confmat(A*w2)

True | Estimated Labels
Labels | 4-clas| Totals
--------|-------|-------
1 | 20 | 20
2 | 20 | 20
3 | 20 | 20
4 | 20 | 20
--------|-------|-------
Totals | 80 | 80

Am I missunderstanding something or this is not working as it should ?

Thanks in advance!
Mariano

[ FIXED ]An error encountered during using Prtool and the relevant book

2008-12-18T07:21:47Z

Hi ,

I have purchased the book

‘Classification,Parameter Estimation and State Estimation--an engineering approach using MATLAB’

. And,I ran the demos of this book - list7.7 & list7.6. I got the listings and data from http://www.prtools.org/book.html . However, some erros were reported by Prtool(4.1). I do not know why.Could I get some help here? The errors and codes are listed here for convenience:

??? Error using ==> setlabtype at 44
Soft labeled datasets should contain at least one class

Error in ==> Listing77 at 2

List 7.6:

z = setlabtype(z,'soft');    % Set probabilistic labels

load nutsbolts_unlabeled;    % Load the data set z
z = setlabtype(z,'soft');    % Set probabilistic labels
[lab,w] = emclust(z,qdc,4);  % Cluster using EM
figure; clf; scatterd(z);
plotm(w,[],0.2:0.2:1);       % Plot results

List 7.7

load nutsbolts_unlabeled;    % Load the data set z
z = setlabtype(z,'soft');    % Set probabilistic labels
                             % Cluster 1D PCAs using EM
[lab,w] = emclust(z,qdc([],[],[],1),4);
figure; clf; scatterd(z);
plotm(w,[],0.2:0.2:1);       % Plot results

Is the tool box useful in handwritten word segmentation, and character recognition

2008-12-12T17:11:31Z

Hi,

I am new to the PRTool.

I would like to know if the tool box is useful in skew detection and correction, handwritten word segmentation, and character recognition.

If yes, please give me tips about which modules could be useful in this so I can go directly to it.

Kind regards,
Maria

A vibration pattern classification example using Prtool & ddtool

2008-12-04T03:04:45Z

FV1,FV3,FV4 are features of the vibration signal of the No.1 No.3 No.4 points.
These 3 points belongs to the same rotary machine but have differnt vibration pattern.
This example shows that our features can discriminate these 3 points.

TrnDataNum1=150;
TrnDataNum2=150;
TrnDataNum3=150;
TrainData=[FV1(1:TrnDataNum1,:)',FV3(1:TrnDataNum2,:)',FV4(1:TrnDataNum3,:)'];
TrainData = dataset(TrainData',genlab([TrnDataNum1 TrnDataNum2 TrnDataNum3],{'No1','No3','No4'}'));
%%
fracrej = 0.02; s =1;
w1 = gauss_dd([],fracrej ,s);
w2 = gauss_dd([],fracrej ,s);
w3 = gauss_dd([],fracrej ,s);
w = multic(TrainData,{w1,w2,w3});

%%
TstDataNum1=50;
TstDataNum2=50;
TstDataNum3=50;
TestData=[FV1(TrnDataNum1+1:TrnDataNum1+TstDataNum1,:)',FV3(TrnDataNum2+1:TrnDataNum2+TstDataNum2,:)',FV4(TrnDataNum3+1:TrnDataNum3+TstDataNum3,:)'];
TestData=TestData';
confmat(getlab(TestData),TestData*w*labeld)

TestData=dataset(TestData,genlab([TstDataNum1 TstDataNum2 TstDataNum3],{'No1','No3','No4'}'));
scatterd([TrainData;TestData],'legend')
plotc(w,'k--');
axis equal

Using Bagging fusion

2008-11-30T13:03:35Z

I just received the question on using baggingc classifier in PRTools by e-mail. As it could be relevant for more people, I post the answer here:

Bagging stands for “Boostrapping and aggregation”. It’s a fusion strategy where a number of simple base classifiers are trained on bootstrapped (sampled with replacement) version of the training data. Eventually, the classifier outputs are combined using voting or mean combiner. The base classifier should be something simple such as decision tree with few levels (stump) or a weakly trained mode (see help of weakc mapping). Due to bootrsapping, bagging does not make much sense for large dataset (where bootstrapped versions of dataset are likely to be almost identical).

PRTools baggingc takes data and untrained mapping to be used as a base classifier. The user must also specify the number of base classifiers to train.

We will use bagging on the banana dataset which can be generated using gendatb command.

>> a=gendatb
Banana Set, 100 by 2 dataset with 2 classes: [50  50]

% our base classifie will be a simpe decision stump with 5 levels
>> wbase=stumpc([],'maxcrit',5)
Decision Stump, untrained  mapping   --> stumpc

% lets perform bagging of 100 decision stumps
>> w=baggingc(a,wbase,100)
Mean combiner, 2 to 2 trained  mapping   --> fixedcc

% lets visualize the classifier output
>> figure; scatterd(a)
>> plotc(w)

The resulting decision boundary nicely separates both classes

About a ‘PRSD License Error’

2008-11-28T07:33:18Z

I have downloaded the evaluation version of PRSD Toolbox 1.1 26-Oct-2008.
And, I have installed it according to the instructions from my registered Email.

During installation, some information were shown in MATLAB command window:

PRSD Studio initialization script
---------------------------------

Searching for PRTools toolbox: PRTools 4.1.4 11-Oct-2008 present

PRSD Toolbox 1.1 26-Oct-2008 present
Searching for the license...not found.
   Please enter the PRSD Toolbox activation key you've received in the download e-mail.
PRSD Toolbox activation key: 7676-1521-2945-9114
Activation successful, the license file written to 'E:\NoInstall\FengKunDevelop\MATLABWORK\Pattern_Recognition\PRSD_Studio_eval_1.1\matlab\prsd_toolbox\prsd_studio_eval.lic'.
The license file was copied to the 'E:\NoInstall\FengKunDevelop\MATLABWORK\Pattern_Recognition\PRSD_Studio_eval_1.1\matlab\..\sdk' directory
PRSD Toolbox 1.1 (26-Oct-2008), Copyright (C) 2008, PR Sys Design, All rights reserved
Evaluation version. Error checking out license: System clock has been set back (-40)
 Retrieve the license using prsd_activate(KEY)
with the KEY you received when downloading the software.
For assistence, contact us at info@prsysdesign.net.

Installation successfully completed.

You can type prsd_demo to see the list of available examples.

However, when I tried ‘prsd_demo(2)’ and clicked the ‘next’ button of the GUI several times,this error infomation was shown:

Error in slideshow.
Error using ==> prsd_mex
PRSD License Error

Is that to say the evaluation version do NOT support the ‘prsd_mex’?

My Environment:

MATLAB 2008b 7.7
MS windows 32bit

PR_Warning: getprior: qdc: No priors found in dataset, class frequencies are used instead

2008-11-28T03:36:19Z

I am a beginner in PR. When I ran the routine(from the book listings) :
______________________________________________________________________

load nutsbolts;
cost = [ -0.20  0.07  0.07   0.07 ; ...
          0.07 -0.15  0.07   0.07 ; ...
          0.07  0.07 -0.05   0.07 ; ...
          0.03  0.03  0.03   0.03 ];
w1 = qdc(z);      % Estimate a single Gaussian per class
w2 = w1*classc*costm([],cost'); % Take cost into account 
scatterd(z); 
plotc(w1);        % Plot without using cost
plotc(w2);        % Plot using cost

______________________________________________________________________

Prtools showed warnning that:
‘PR_Warning: getprior: qdc: No priors found in dataset, class frequencies are used instead’

why did this warnning occur? And, how to add priors in dataset?
I am puzzled.

My PC environment:
MATLAB 2008b ,Ver :7.7
Prtools 4.1 academic
OS: Windows XP professional

Stable and development branches

2008-11-10T06:15:58Z

In order to reflect different goals of our customers, we have decided to split the PRSD Studio into two branches, one stable and the other development. The development one will continue to add new features and gradually stabilize eventually leading to the new stable release intended for production. The stable branch will only fix bugs found in the stabilization process. The version numbering of the stable branch uses two numbers such as 1.1 or 1.2. The development releases will add the third step (e.g. 1.1.1 or 1.1.2). The development in the 1.1.x line will lead to the new stable release 1.2. Each version number is accompanied with the date (1.1 04-Nov-2008)

The release notes of each public release are available in the documentation. For the development releases, we also include quick examples so you can directly try out the new features.

The latest development version (1.1.1 09-Nov-2008) adds leave-one-out cross-validation of arbitrary algorithms and feature selection performed in an inner cross-validation loop.

Best Regards,

Pavel

Hi from Carmen

2008-11-08T11:59:25Z

Hallo everybody!

I’m an italian happily living in The Netherlands. I’ve been applying pattern recognition techniques to genetic research on cancer during my PhD project pursued in the Bioinformatic group at TUDelft. During this project I enjoyed the collaboration with the researches at the Dutch Cancer Institute.

I have recently joined Pavel in PR Sys Design , enjoying lecturing in the APR courses and working on our software package.

enjoy your days :)
Carmen