unsupervised contrastive loss

[Dante-Basile]

Reference issue

#520

Type of change

This contains the loss function and a basic ResNet50 demo demonstrating training and decreasing loss on CIFAR-10.

What does this implement/fix?

This demonstrates that the ResNet50 network can effectively optimize the loss of the unsupervised contrastive loss function. Future experiments will examine the accuracy of the trained network.

Additional information

NDD 2021

[PSSF23]

Refer to file structure in #518

[codecov]

Codecov Report

Merging #522 (8373149) into staging (634d4d1) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           staging     #522   +/-   ##
========================================
  Coverage    90.09%   90.09%           
========================================
  Files            7        7           
  Lines          404      404           
========================================
  Hits           364      364           
  Misses          40       40           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 634d4d1...8373149. Read the comment docs.

[PSSF23]

@Dante-Basile You need to black format the files.

@jdey4 do you think this is an ideal way to separate supervised & unsupervised? Or should there be separate folders in benchmarks?

[PSSF23]

• Your notebooks have unnecessary outputs & empty cells. Remove them.
• Remove commented lines of code