0
$\begingroup$

Are there any architectures that allow two partners to build a model while protecting each of their respective intellectual property (IP)?

For example:
Company A builds machine learning models
Company B has a large data set

Company A does not want to expose how it builds their models, and company B can not loose control of their (HIPPA/medical) data.

Thank you in advance for any thoughts you might have on this.

$\endgroup$
2
  • $\begingroup$ What does the data look like? Is it all numeric? You say HIPPA so I’m guessing there is string data too? $\endgroup$ – kbrose Jul 19 '18 at 13:47
  • $\begingroup$ @kbrose: I think it will be a mix of data types certainly including strings and numeric values. The domain is medical diagnostics so heart rates, blood pressure, medicines, treatment times, etc. $\endgroup$ – Ryan Lowe Jul 19 '18 at 18:14
0
$\begingroup$

I understand that company A would be required to view the data, which is IP of company B. Not sure why company A would expose the model to company B. For the former, they could rename the variables and maybe even transform them (creating pseudo data). For the latter, well, they just don’t say/show anything :) Hope this helps

$\endgroup$
1
  • $\begingroup$ Thanks @Srikrishna. I would agree with you that traditionally company A would have to view the data of company B. Non-disclosure agreements would be used to provide legal protection and so forth. I am hunting for alternative architectures that somehow circumvent the usual way of partnering and building a model. $\endgroup$ – Ryan Lowe Jul 18 '18 at 18:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.