News for hackers. Stuff that matters.

A comment on Diffusion Models Beat GANs on Image Synthesis.

Re:Diffusion Models Beat GANs on Image Synthesis by ansk on Tuesday May 11, 2021 @11:06PM (#27126149)
Why broadly claim an entire class of generative model outperforms another? They simply show that their particular implementation of a diffusion model attains better performance (for some metric) than the current best implementation of a GAN. Almost inevitably (and likely in the very near future) someone will invest the time and tuning to squeeze out some more performance from a GAN/VAE/flow, and this title will be outdated.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by world_peace42 on Wednesday May 12, 2021 @12:06AM (#27126477)
Red Sox beat Yankees in the world series =\= Red Sox will always beat the Yankees in every world series game. And I’m guessing most people reading this title know that. I’m sure there are people new to the field or new to CS research in general who might not know how to interpret this, but the only thing most people take out of this title is “hmm, maybe we should consider and study diffusion models more seriously”.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by marksmith2996 on Wednesday May 12, 2021 @01:16AM (#27126863)
The Red Sox and Yankees can't play each other in the World Series fyi.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by ansk on Wednesday May 12, 2021 @12:23AM (#27126561)
The preeminence of either GANs or diffusion models is an ongoing debate, with evidence trickling in paper by paper. To claim diffusion models beat GANs is like claiming the Red Sox beat the Yankees because they happen to be ahead in the third inning.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by catillac on Wednesday May 12, 2021 @12:34AM (#27126623)
It’s like saying the Sox are ahead of the Yankees and it’s the third inning with no predetermined number of innings. That nuance is what makes the title fair.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by bearzoo on Wednesday May 12, 2021 @12:04AM (#27126465)
Eh I feel like may be worth cutting some slack here..

It sounds like you’d like the title to be prefaced with “one time that”.

I don’t think anyone reading this says, hey gans are done for. I doubt the authors think that and I (clearly) don’t get the impression that’s what they mean.

Even the first sentence of the abstract:

> We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models

Has the word current in front of state of the art generative models.

[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by jostmey on Wednesday May 12, 2021 @12:34AM (#27126618)
Why I agree with your sentiment, GANs always seemed finickily and unreliable. I am willing to indulge new approaches to find something more robust
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by eru on Tuesday May 11, 2021 @11:31PM (#27126290)
While I tend to agree with you, aren't title like these understood to be short hand for what you described?
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by forgingahead on Tuesday May 11, 2021 @11:42PM (#27126343)
Diffusion Models Can Sometimes Beat GANs on Image Synthesis
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by ansk on Tuesday May 11, 2021 @11:47PM (#27126368)
For an actual example of this more cautious framing, see another paper (also by OpenAI, also comparing two classes of generative models), https://arxiv.org/abs/2011.10650.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by sillysaurusx on Wednesday May 12, 2021 @01:41AM (#27127007)
It's fine to claim it. It's shorthand for "We achieve smaller FID values (most of the time) than any other generative model."

You're correct about what you're saying, though. It's also worse: FID is a measurement biased towards imagenet-style images, and this model was also trained on imagenet, so it's quite well-suited for achieving small FID values. Whereas for generative anime, it's still unclear what the best model architecture is.

[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by bigyikes on Tuesday May 11, 2021 @11:47PM (#27126369)
I read your comment, thought “to get clicks, duh”, and then I clicked through the link.

This is a research paper.

I would like to echo your question: why? Does academia incentivize clickbait?

[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by Fomite on Wednesday May 12, 2021 @12:14AM (#27126517)
With the rise of Altmetrics? Arguably yes.

IIRC, the last research I saw on this, which was awhile back, argued that clever titles get read more and cited less. https://www.researchtrends.com/issue24-september-2011/headin...

[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by eru on Wednesday May 12, 2021 @12:28AM (#27126585)
Well, not sure it's any worse than cite-bait?
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by peteretep on Wednesday May 12, 2021 @12:35AM (#27126626)
Are you going on the title? If so, the verb "beat" here admits both a singular past action and a present continuous action.
[ Parent | on HN ]