forward_transform about compression

[jonycgn]

Hi Serhan,

you are correct. The forward and backward transforms are currently left as a pseudocode placeholder. I'll take this as a suggestion to flesh out the example a bit more. We'll likely do this when we get to expanding the library a bit more (should happen soon-ish).

You can just define any appropriate CNN. As a tip, it generally helps to drop the activation function before the bottleneck, and not to use batchnorm!

Best!

[jonycgn]

Serhan, we've added a lot of new code in the last few days, including a fully working example, which also defines the transforms. Please check it out and let us know about any issues.