Home>

Tell me about licenses and copyrights related to deep learning.

Neural network weights may be licensed,
Is there any license or copyright for the configuration of the neural network, and is it commercially available?

For example, if you learned the same network configuration such as yolo and Vgg from 1 with original data, is it commercially available?

  • Answer # 1

    https://softwareengineering.stackexchange.com/questions/319358/signification-of-yolo-license

    https://github.com/huyng/tensorflow-vgg/blob/master/LICENSE

    https://opensource.stackexchange.com/questions/6961/how-does-gpl-apply-to-neural-network-models

    I think it's difficult to license the network structure itself.

    or when the weights published as text are used as initial values.

    If you want to go outside for internal use, it is better to have a proper specialist scrutinize it.


    Unrelated chat

    Maybe the structure of the neural network will become obsolete and will shift to an automatically generated model that deals well with specific problems.

    Continuing to cultivate data scientists who are not sure whether the unit price can be high or not is a heavy burden, and if parameter tuning can be done automatically, it seems that it will not exceed it.

    Don't be afraid because there is nothing else that data scientists think about in general.


    A story that was a little interesting when I examined it
    https://www.news.com.au/technology/online/piracy/possession-of-this-illegal-prime-number-could-get-you-arrested-in-america/news-story/bf104dd402a75e2da08d97e66d28e4df

    I can't predict where things will get caught.

  • Answer # 2

    https://github.com/huyng/tensorflow-vgg/blob/master/LICENSE

    Looking at this, this source is written for commercial use as Permissions.
    Rather, MIT and the open source system thought that it would be commercially available if the source code was released, but there seems to be a subtlety

    Permissions

    Commercial use
    Modification
    Distribution
    Private use