Dear Tom:

I agree - this is another puzzling and troublesome part of Section 21.  Why should any competitor be prevented from publishing the results of their work?  The rights exclusively granted to HPN include, "... publicly perform, publicly display, digitally perform", and the "Licensed Materials" include the predictive algorithm.  Even if you use a completely different data set, they still control virtually all of the rights to your work, including the right to publish.

I plan to bring this up with Dr. Merkin, in addition to the ownership / IP issues.

Mark

To Competitors:  The HPN 'Click Through' Agreement is Probably Enforceable

A number of competitors have questioned whether the 'click through' agreement on the HPN Health Prize / Kaggle site is actually enforeable if you don't win a prize.  As noted in my posts in the "Boycott Heritage Health Prize" forum, Section 21 of this agreement grants HPN extensive rights to your work -- the minute to "click through" to register.

For a concise summary of click through agreements suitable for people who are not attorneys, see:

     http://en.wikipedia.org/wiki/Clickwrap

A number of relevent cases are listed.  This applies primarily to the United States, of course.  Laws vary in other countries.

Mark Hays

To all current and potential Health Prize competitors: 

I sent a letter via FedX to Dr. Richard Merkin, CEO of HPN, regarding the concerns around Section 21 of the online Competition agreement.

Hopefully we will be able to resolve these issues, to strengthen and ensure the success of the Competition -- which could improve healthcare worldwide.

Mark Hays

@Tom SF Haines, Jeremy is not the author of the rules. He is merely trying his best to point people to the section that makes the rules as competitor friendly as possible given HPN's requirements.

Also, if you would like to publish your algorithm, I strongly encourage you to put in a research request using the Contact Us form.

Mark Hays: Did you ever get a response to your letter? I would of thought 17 days would be plenty of time, even when dealing with an old fashioned postal system!

So far I am planning to use MatLAB as a 'tool' to develop my algorithm, because I can setup and test the code quickly and see if my results are competitive or not.

When everything is trained and running, the run-time implementation of the neural network can be written in GCC C, Octave, etc.

Look forward to the organizers to further refine the rules on tools. I would think HPN would prefer the best performing algorithm developed with industry standard tools that are suppported, rather than something that doesn't work as well but is open-source and written in a lower level language.

Kurt
Salt Lake City, UT

@Dobson - using MatLAB is fine.