add selu activation function

[lakshayg]

resolves #10888

This pullrequest changes

• adds the selu activation function

I have not yet added the test data and request the devs to guide me on how to do it.

[dkurt]

Could you please provide an origin of these magic values? Are they results of some optimization problem that we can reproduce locally or they are just some irrational numbers that have too complicated formula? This way it'd be nice to put a reference to a some proof that they are optimal.

[lakshayg]

I have put a link to the paper in the linked issue. Here it is again: https://arxiv.org/abs/1706.02515

[dkurt]

@lakshayg, thank you for your contribution! Unfortunately, I did not find any strong proofs why this activation is so useful. I think we can add this layer later, when it'll be more popular. For now you can test custom layers creation from PR #11129.

[lakshayg]

Sure, no problem.

[dkurt]

Please look at custom layers registration: https://docs.opencv.org/master/dc/db1/tutorial_dnn_custom_layers.html.

[AlvarHHM]

@lakshayg, thank you for your contribution! Unfortunately, I did not find any strong proofs why this activation is so useful. I think we can add this layer later, when it'll be more popular. For now you can test custom layers creation from PR #11129.

Both pytorch, keras and tensorflow has native support for SELU, which is a strong evident that this activation layer should be popular enough for opencv to support it.

P.S. I would be grateful if someone can provide a custom layer implementation.