Challenge 4: code submission

Submitted by ACGT on Sun, 2011-07-17 12:53

Hey,

In challenge 4 there's a requirement to submit the code: "We also request the submission of a file containing the code required for reproducing your predictions".

Our question is which code is required: Is it the code that get the training data and learns a model that enable it to predict the test data, or the code that receives a test data and predicts it (while the model was already learned and is hard-coded in the code)?

If it is the first case, then this is problematic. In many cases, the learning algorithms are used via other tools and even commercial tools. It is problematic to create a self-contained code, when it is implemented in another tool, and in case of commercial tools the license agreement might not permit it.

We believe it is better to give the pseudo-code, as is requested in the other challenges.

Thanks, team ACGT

‹ Parameter Estimation challenge: dissociation constant and Hill coefficient General Challenge Submission ›

Comments

Code Sumnission issue for Challenge 4

Submitted by Dream_manager on Sun, 2011-07-17 23:27.

Hi ACGT

good question. We are requesting the code that receives a test data and predicts it (while the model was already learned and is hard-coded in the code). In the write up, we are requesting the pseudo code for the creation of the rules for classification from training data.

Is this a reasonable request, given the constraints that you list?

Gustavo 

code for prediction

Submitted by Admire-LVQ on Wed, 2011-08-10 08:29.

In which format should the code "that predicts a test data"  process the data?  By reading in all files associated with

a single patient (number specified as input)?

Or should it delive the predictions for all 180 test cases in one go?

A little more detailed description of the required code would be helpful.

Thanks in advance,

Michael
 

Hi Michael, thanks for your

Submitted by Dream_manager on Thu, 2011-08-11 11:03.

Hi Michael, thanks for your question

You should explain for a single patient how you decided its phenotype.

thanks

Pablo

code submission

Submitted by Admire-LVQ on Fri, 2011-08-12 11:31.

Dear Pablo,

thanks a lot for your reply. Of course we will explain our method

in the submitted material. My question, however, was about code submission.

In your earlier reply, it is said that code should be provided

that classifies the test cases. If this is still required, I think we would need

to know more details. How should the data be read in, for instance?

Best regards,

Michael

No Code Will Be Required for Challenge 1 submission

Submitted by Dream_manager on Sun, 2011-08-14 04:23.

Michael,

Following some suggestions by other participants, we will not require code submission.

Thanks!

 

Gustavo