An analytic theory of creativity in convolutional diffusion models · HackerPapers discussions

An analytic theory of creativity in convolutional diffusion models — HN submission

Mon, 11 Aug 2025 17:52:34 GMT

An analytic theory of creativity in convolutional diffusion models

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

An analytic theory of creativity in convolutional diffusion models — mentioned by yorwba

Thu, 17 Jul 2025 20:20:33 GMT

It's more likely that such an architecture would be bigger rather than smaller. https://arxiv.org/abs/2412.20292 demonstrated that score-matching diffusion models approximate a process that combines patches from different training images. To build a model that makes use of this fact, all you need to do is look up the right patch in the training data. Of course a model the size of its training data would typically be rather unwieldy to use. If you want something smaller, we're back to approximations created by training the old-fashioned way.

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

An analytic theory of creativity in convolutional diffusion models — HN submission

Wed, 09 Jul 2025 20:46:28 GMT

An analytic theory of creativity in convolutional diffusion models

7 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

An analytic theory of creativity in convolutional diffusion models — HN submission

Tue, 01 Jul 2025 21:30:04 GMT

An analytic theory of creativity in convolutional diffusion models

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

An analytic theory of creativity in convolutional diffusion models — mentioned by Retr0id

Tue, 01 Jul 2025 12:24:16 GMT

The paper: https://arxiv.org/pdf/2412.20292

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

An analytic theory of creativity in convolutional diffusion models — HN submission

Mon, 06 Jan 2025 10:12:56 GMT

An analytic theory of creativity in convolutional diffusion models

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.