I would like to bet against this statement. What are the stakes?
I bet linear regression predicts global temperature better than climate models, and without the months of supercomputer time.
-
-
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
You're making a good point there, don't know about the specific statement but the simplicity of models in general is under-appreciated. It's not science to throw raw compute power at something, we should aspire to develop more simpler models, and keep simplifying existing models.
-
Complex models do a lot of overfitting. Restricting the complexity of models to some degree was the basic principle till a few years ago.
- Näytä vastaukset
Uusi keskustelu -
-
-
Sshhh… quiet
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
This would be a cool experiment, although the reason those models are slow is because they’re modeling a multi-part system we call climate (oceans, gases, water vapor, albedo, etc.) at a resolution of, say, 50x50 km. Rather than just temp.
-
Yes, but from what we know of ML, that's probably not the best way to predict the temperature. (Or: don't solve a problem by way of a harder problem.)
- Näytä vastaukset
Uusi keskustelu -
-
-
I would be interested to see this experiment done.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
You're putting down Navier's and Stoke's contributions to physics.
-
No, the point is that, based on what we know of ML, Navier-Stokes is probably far from the best way to predict global temperatures.
Keskustelun loppu
Uusi keskustelu -
-
-
Is that a fact or an opinion?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.