Hello, Lumin a écrit : > [...] > Even if upstream releases their pretrained model under GPL license, > the freedom to modify, research, reproduce the neural networks, > especially "very deep" neural networks is de facto controled by > PROPRIETARIES.
In my opinion, a company releasing a pretrained neural network model under GPL License may be taking a legal risk. This can be an interesting case for lawiers. Are some of us, Debian fellows, lawiers? I am no lawier; my father was lawier then judge, my sister is judge, but she is seldom concerned with such cases or approaching cases. Let us suppose that company X released a pretrained model Y and released it under GPL license. Now, as a genuine end-user, I can download Y, and read its license. Oh! it is GPL! so, I can ask for sources of Y. The question is, what are sources of Y? I presume that any sane judge would agree that in that case, sources are the set of materials and immaterials which allow one to rebuild consistently the object Y. Hence, the sources are: 1- the random matrix used to initialize the neural network 2- the training algorithm 3- the (big) data to feed the training process 4- if no commercial machine exists at a "fair price", to train the neural network in a reasonable amount of time, the publisher of Y must lend a machine to fulfill the GPL requirements. That is, in my opinion, what GPL does enforce if we sue the company X for infringement of the license. Do you think it is worth a try? Do you know a friend organization which would like to involve in such a trial? Best regards, Georges.
signature.asc
Description: PGP signature