SmoothMix Addition

[asarj]

Description

• Added support for SmoothMix training on Randomized Smoothing classifiers.
• Created a separate estimator for SmoothMix training (as per Addition of SmoothMix Training in the Randomized Smoothing Module #1668 (comment))
• Migrated the training procedure from PR Addition of SmoothMix Training in the Randomized Smoothing Module #1668 into the fit method of the SmoothMix estimator

Paper Link: https://arxiv.org/pdf/2111.09277.pdf
Code Repository based on paper: https://github.com/jh-jeong/smoothmix

Fixes # (issue)

Type of change

Please check all relevant options.

• Improvement (non-breaking)
• Bug fix (non-breaking)
• New feature (non-breaking)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Testing

Please describe the tests that you ran to verify your changes. Consider listing any relevant details of your test configuration.

• Test A
• Test B

Test Configuration:

• OS: Ubuntu 22.04 LTS
• Python version: 3.7
• ART version or commit number: 1.11.0
• TensorFlow / Keras / PyTorch / MXNet version: PyTorch 1.11.0+cu113

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[asarj]

Hi @beat-buesser, created this new PR to reflect your recent suggestions from #1668 (comment) and to also avoid the git issues from the previous PR. I am currently looking into implementing the SmoothMix authors' SmoothPGD implementation inheriting from art.atttacks.attack.EvasionAttack, but will be traveling soon and I'm not sure if I will be finished with this part before then, will update this draft accordingly.

Tagging @keykholt and @rahmati for updates, and will close PR #1668. Please let me know if you have any other questions in the meanwhile

[lgtm-com]

This pull request introduces 3 alerts when merging 459f209 into 8df19af - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[codecov-commenter]

Codecov Report

Merging #1705 (c366b1d) into dev_1.13.0 (8df19af) will decrease coverage by 33.22%.
The diff coverage is 53.10%.

❗ Current head c366b1d differs from pull request most recent head c415a63. Consider uploading reports for the commit c415a63 to get more accurate results

@@               Coverage Diff               @@
##           dev_1.13.0    #1705       +/-   ##
===============================================
- Coverage       88.09%   54.87%   -33.23%     
===============================================
  Files             259      251        -8     
  Lines           21347    23464     +2117     
  Branches         3789     3942      +153     
===============================================
- Hits            18805    12875     -5930     
- Misses           1598     9829     +8231     
+ Partials          944      760      -184     

Impacted Files	Coverage Δ
...cks/evasion/adversarial_patch/adversarial_patch.py	55.88% <0.00%> (-1.70%)	⬇️
...adversarial_texture/adversarial_texture_pytorch.py	12.50% <0.00%> (-77.45%)	⬇️
...ion/imperceptible_asr/imperceptible_asr_pytorch.py	7.86% <0.00%> (-78.94%)	⬇️
..._air_flickering/over_the_air_flickering_pytorch.py	17.91% <0.00%> (-76.87%)	⬇️
art/attacks/evasion/pixel_threshold.py	19.80% <0.00%> (-71.50%)	⬇️
...ted_gradient_descent/projected_gradient_descent.py	83.78% <ø> (-13.52%)	⬇️
art/attacks/evasion/saliency_map.py	22.89% <0.00%> (-77.11%)	⬇️
art/attacks/evasion/universal_perturbation.py	27.10% <0.00%> (-65.43%)	⬇️
art/attacks/evasion/wasserstein.py	11.94% <0.00%> (-83.28%)	⬇️
art/attacks/evasion/zoo.py	12.01% <0.00%> (-71.71%)	⬇️
... and 240 more

[asarj]

Marking for review, but note that I did not implement the SmoothPGD attacker by inheriting from the EvasionAttack module (will be away for some time so not sure when I'll be able to resume working on this), but I welcome feedback otherwise

[lgtm-com]

This pull request introduces 3 alerts when merging a2ebed6 into 8df19af - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[lgtm-com]

This pull request introduces 3 alerts when merging c415a63 into 3f106d7 - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[lgtm-com]

This pull request introduces 3 alerts when merging c415a63 into 89bf92f - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[lgtm-com]

This pull request introduces 3 alerts when merging c415a63 into c366b1d - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[lgtm-com]

This pull request introduces 3 alerts when merging c415a63 into 6ab005c - view on LGTM.com

new alerts:

• 3 for Module is imported more than once

[beat-buesser]

Closing following #1988