Hi

I'm getting confused by the laws so I am going to ask questions that are relevant to me that will just require YES or NO answers

 

1. Can I use R?

2. Can I use Weka?

3. Can I use Excel?

4. If I organise the data in a novel way and just use a standard processing algorithm, such as Naive Bayes, is this OK?

 

Many thanks, this is all I need to know

Hi @Domcastro.

1. Can I use R?

Yes

2. Can I use Weka?

Yes

3. Can I use Excel?

Yes

4. If I organise the data in a novel way and just use a standard processing algorithm, such as Naive Bayes, is this OK?

Yes

The bit about using tools requiring licensing requiring prior approval is confusing me...since Excel does require a license, so i'm wondering, can we also use non-open source software for modeling such as SPSS or SAS?  if so does the requirement to provide "code" along with documentation refer to SPSS or SAS code for any part of the process involving these tools?  

Yes, you can use Excel to create your algorithm - the licensing issue only is relevent when considering what is required for HPN to actually implement your algorithm after the close of the competition. In general, models created in Excel can normally be implemented in production without requiring Excel (e.g. by implementing the final algorithm in some programming language like C++ etc). However if implementation of your model by HPN will require a license to some tool/library/etc, then you will need to get permission from HPN.

Of course, for something simple like Excel, HPN will almost certainly give that permission, since if they don't already have it they can easily purchase that license themselves. The only situations I expect you would see HPN deny permission would be in cases where obtaining and maintaining a required license may involve risks to HPN (e.g. software libraries that don't allow commercial use).

Thank you.  I would like to request permission to use SPSS, SPSS PASW Modeler, Excel, MS SQL, R and PERL.  Could somebody please point me to where to submit this request or see where I could check to see what software has been approved already?

I believe Perl and R are under OSI approved licenses, so no problems there. I'll ask someone from HPN to get back to you regarding the remaining.

Awesome, thanks.  Would it make sense and be possible to keep a running list somewhere that folks can reference that lists approved software, so that it only needs to be requested one time and it's very clear about using even R and PERL...since there have been so many questions on this and i think it's difficult to find buried in the forum...?

@jwdatagirl: Yes absolutely - I've asked HPN is this is OK. I agree it's a very good way to handle it.

As I did not found not yet any running list of approved software (which would be a good thing to have), can you confirm it is allowed to use SAS?

Confirmed.

Hi,

I would like to be absolutely sure that what I am using is not going to raise issues. Can you please confirm (or refute) that these software packages are allowed?

• CART, MARS from Salford Systems - some of the ensemble models from CART require CART for running the model, others can be run in any language
• SAS Enterprise Miner - Neural Network node. The trained model may be executed in Base SAS 
• GenIQ - this is a little known software. The models produced by this software are written in plain equations that may be written and executed in any basic language
• Matlab Neural Nets and SVM - to run the NN or SVM models you do need to have Matlab

Thanks.

To clarify the rules (based on rule 21 and the definition of "prediction algorithm" earlier in the rules):

Any software required to execute the prediction algorithm must fall into one of three categories: (a) you have unrestricted right to grant a license to HHN, (b) must be in the list of OSI-approved licenses (http://opensource.org/licenses/alphabetical), or (c) must be software for which HHP has waived these requirements.

Based on above written approval, Matlab & Base SAS fall under (c), so that's fine. You said that Base SAS is all that's required to execute the models. Can you confirm that only Matlab and no add-ons are requried for those models?

Unless you produce code that would execute the models without CART/MARS & Gen software, those would need to be exceptions. I'm looking into licensing difficulty/costs for those, which is what would determine whether HHP is willing to make an exception for them.

Can you also confirm

SAS/Stat
SPSS
?

I am greatly enjoying reading over the winners' techniques, and now wondering what if any of this I can share with my professional collegues?  Can the 'how I did it' papers be shared outside of the competition participants, or would that be an infringement of copywrite, or some specific rule of the competition?  I did read over the rules and terms but I don't see anything that specifically addresses this question.

Safest thing would be to give your colleagues a link to the papers rather than making a copy. They don't need to be registered for the contest for the link to work.