Added new flexible loss named GeneralLoss

[ashutosh1919]

fixes #1646.
@gabrieldemarmiesse , @pavithrasv and @mihaimaruseac, please review.

[WindQAQ]

Hi @ashutosh1919 , thanks for the PR. Because this PR is somewhat large, let's improve it step by step :-)

For testing, please follow other tests in loss module.
https://github.com/tensorflow/addons/blob/master/tensorflow_addons/losses/tests/kappa_loss_test.py

[gabrieldemarmiesse]

As a side note, maybe we should reconsider the name "GeneralLoss"? It may be confusing for users. They may think it's some kind of abstract base class.

We also changed the CI quite a bit recently, so I'm merging master to make sure everything works. Please do git pull before adding new commits.

[ashutosh1919]

@gabrieldemarmiesse , Yes you are right. We should think about better name here. GeneralLoss is quite confusing.
So, @gabrieldemarmiesse and @WindQAQ , please suggest name if you have any in your mind. I will also search for it.

[ashutosh1919]

@gabrieldemarmiesse , @WindQAQ - How about the name UniversalLoss (universal_loss) ?

[AakashKumarNain]

Not a good one IMO

[gabrieldemarmiesse]

There are only two hard things in Computer Science: cache invalidation and naming things

• A famous dude

Cache invalidation is easier IMO

[ashutosh1919]

Some suggestions for name:

• generalized_gaussian_loss (from research paper, it is noted that the behavior of this loss is similar as Generalized Gaussian Distribution.)
• cauchy_gaussian_loss (It is also mentioned in the paper that this loss covers the distribution which is intersection of cauchy and gaussian distributions.)
• general_robust_loss (general_loss may confuse developers but this one will not. And btw, this loss is one among the robust_loss implemented by google research.)

[ashutosh1919]

@gabrieldemarmiesse , Any update on this PR? Thanks!

[bhack]

gar_loss (General and Adaptive Robust Loss)

[ashutosh1919]

gar_loss (General and Adaptive Robust Loss)

@bhack,
Don't you think adaptive_loss is different than general_loss as per the code here.

[ashutosh1919]

@gabrieldemarmiesse , @bhack , @AakashKumarNain - Any suggestion for name to go ahead with this PR?

[bhack]

It don't know what to do exactly with all these Docs repos.
We have focal also in https://github.com/keras-team/keras-cv

[ashutosh1919]

@bhack , I think you misunderstood this PR with #1732 . That PR was focusing on doc tutorials for Focal loss & Contrastive loss.

This PR adds a new loss General loss.

[bhack]

Yes I was fooled by github notifications

[seanpmorgan]

Thank you for your contribution. We sincerely apologize for any delay in reviewing, but TensorFlow Addons is transitioning to a minimal maintenance and release mode. New features will not be added to this repository. For more information, please see our public messaging on this decision:
TensorFlow Addons Wind Down

Please consider sending feature requests / contributions to other repositories in the TF community with a similar charters to TFA:
Keras
Keras-CV
Keras-NLP