Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations — mentioned by johnsutor

Thu, 22 Feb 2024 16:38:30 GMT

This is the premise behind some more recent self-supervised learning papers (see https://arxiv.org/abs/2303.03307). Turns out, it's a solid analog for classification tasks.

Paper and discussion history. A mention is not an endorsement.

Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations · HackerPapers discussions

Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations — mentioned by johnsutor