Welcome to RSCTC'2010 Discovery Challenge

Questions, answers, discussions related to RSCTC'2010 Discovery Challenge

Welcome to RSCTC'2010 Discovery Challenge

Postby RSCTC » Tue Dec 01, 2009 12:26 pm

Welcome Everybody!

Thank you for your interest and participation in RSCTC'2010 Discovery Challenge. We hope you'll enjoy the contest and have great fun struggling with one of the most difficult and important problems at the frontiers of data mining, bioinformatics and genetics. Your new fresh ideas may bring us closer to efficient exploitation of genetics information in medicine and pharmacology, for saving lives of millions of people suffering from serious diseases, like tumors, heart failures and others.

If you want to discuss any issues related to the challenge, please post on this forum. We'll do our best to answer all questions.

Have a good fun!

Organizing Committee
RSCTC
 
Posts: 1
Joined: Thu Oct 29, 2009 6:44 pm

Re: Welcome to RSCTC'2010 Discovery Challenge

Postby harriers » Wed Dec 02, 2009 11:45 pm

hi, was wondering about couple of things:

the train files in data[x]-train.zip have class label while test files don't
if the six data[x]-test files are not the official=secret test sets
(i.e. if they are "example / public datasets" as it says)
how are we to evaluate our algorithms here on basic track
(is it by cross-validating on the data[x]-train files?)

I'm more interested in basic track, with no jar compiling
it says "in basic track the participant submits a text file
with predicted decisions for test samples"
yet on Submit site I only find option for evaluating the algorithm
on the advanced track (jar)

best wishes, Harri
harriers
 
Posts: 5
Joined: Wed Dec 02, 2009 9:20 pm

Re: Welcome to RSCTC'2010 Discovery Challenge

Postby Winsteps » Thu Dec 03, 2009 4:02 am

A really dumb question:
Is the task to classify the rows of each test dataset into the decision-classes defined by the rows of the matching training dataset (with each training-test pair independent of the others), and then to report the decision classes for the 6 test datasets in 6 submission files (zipped, etc. together)?
Winsteps
 
Posts: 3
Joined: Thu Dec 03, 2009 2:52 am

Re: Welcome to RSCTC'2010 Discovery Challenge

Postby Marcin » Thu Dec 03, 2009 1:28 pm

Winsteps wrote:A really dumb question:
Is the task to classify the rows of each test dataset into the decision-classes defined by the rows of the matching training dataset (with each training-test pair independent of the others), and then to report the decision classes for the 6 test datasets in 6 submission files (zipped, etc. together)?

Exactly. This is the task on Basic track. For each pair, train an algorithm on training dataset and use it to classify rows of test dataset. Each pair is independent of the others, so each pair may have different number of possible classes.

On Advanced track, the task is to implement and submit the algorithm itself, instead of classification decisions.

-Marcin
Marcin
 
Posts: 115
Joined: Fri Oct 09, 2009 6:45 pm

Re: Welcome to RSCTC'2010 Discovery Challenge

Postby Marcin » Thu Dec 03, 2009 1:54 pm

harriers wrote:the train files in data[x]-train.zip have class label while test files don't
if the six data[x]-test files are not the official=secret test sets
(i.e. if they are "example / public datasets" as it says)
how are we to evaluate our algorithms here on basic track
(is it by cross-validating on the data[x]-train files?)

I'm not sure if you didn't miss the page with Basic task description? Each track has its own task description, see: task on Basic track, task on Advanced track. On Basic track, only decisions of test samples are secret, not the whole datasets (as on Advanced). Participants should predict decisions for test samples and submit them as a solution. They will be compared with the secret ground-truth decisions to calculate the score.

harriers wrote:how are we to evaluate our algorithms here on basic track
(is it by cross-validating on the data[x]-train files?)

Do you mean your own evaluation of algorithms, before submission of a solution? If so, you're right, you may cross-validate on the data[x]-train files.

harriers wrote:I'm more interested in basic track, with no jar compiling
it says "in basic track the participant submits a text file
with predicted decisions for test samples"
yet on Submit site I only find option for evaluating the algorithm
on the advanced track (jar)

Don't worry about the "Solution class name ..." field on Submit form. On Basic track just ignore it, you submit a .zip file and so you shouldn't type anything in the class field.

Regards, Marcin
Marcin
 
Posts: 115
Joined: Fri Oct 09, 2009 6:45 pm

Re: Welcome to RSCTC'2010 Discovery Challenge

Postby Winsteps » Fri Dec 04, 2009 12:35 am

Thank you, Marcin. Really helpful replies.
Winsteps
 
Posts: 3
Joined: Thu Dec 03, 2009 2:52 am


Return to RSCTC'2010 Discovery Challenge

Who is online

Users browsing this forum: No registered users and 1 guest

cron