Alright :D
This is wild! @MCHammer retweeted our work on Fourier neural operator with @ZongyiLiCaltech @kazizzad it can solve Navier Stokes and other PDEs from scratch. Thanks @MCHammer for using your influence for popularizing science #scicomm #AI #DeepLearninghttps://twitter.com/MCHammer/status/1319359558070857728 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
My papers.. cant touch that
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Neat. Was this done in FP64? Would the convolution in the middle scale well to lower precision types, maybe even small ints?
-
Thanks! The experiments are done in default float32. We haven't tried low-precision training/inference yet.
End of conversation
New conversation -
-
-
Can't differentiate this.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Too legit to quit expanding our understanding. It's always been Hammer Time, in my book.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Great! I wonder, do you think this approach could be used to speed up training of very large models, by training at a lower resolution?
-
Yes and No: yes the model can be trained on the lower resolution to speed up training, but the lower resolution input shall be accurate (e.g downsampled). It means we still need to generate the data with high accuracy which can be slow.
End of conversation
New conversation -
-
Show additional replies, including those that may contain offensive content
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.