3 ms·
Transformers do this for any stream of tokens, those tokens can map to anything you want, and you will get lossy compression. Text produced by humans just happe
by energy123 4mo ago
Transformers do this for any stream of tokens, those tokens can map to anything you want, and you will get lossy compression. Text produced by humans just happens to be dense, available, and a useful prior, but it is not intrinsically required. See 3D vision transformers for example.
- jeremyjh 4mo agoIt is not possible to compress arbitrary data. If the data is already compressed, or it is encrypted, or it is randomly generated, it cannot be compressed with any method. This is foundational information theory. https://en.wikipedia.org/wiki/Lossless_compression#Limitations https://en.wikipedia.org/wiki/Lossless_compression#Limitatio...
- eru 4mo agoEh, you can't compress truly arbitrary data, sure. But you can compress almost any natural occurring data: reality is very sparse. For some reason.
- thadt 4mo agoWhereas if we're talking about lossy compression (as is the person to whom you replied) we certainly can compress arbitrary data - almost as much as we want. The hard question, then, is how much the decompressed output looks like the original.