Ghosts of Softmax: Zeros of the partition function explain training instability

Posted by g4omingron on Tuesday March 17, 2026 @02:16AM
from the github.com dept.

https://github.com/piyush314/ghosts-of-softmax

( 2 comments | on HN )

[dead]

Re:Ghosts of Softmax: Zeros of the partition function explai... by yorwba on Tuesday March 17, 2026 @04:40AM (#47410016)

Nice work! The paper feels verbose at times and could use some editing to slim it down (also, equation 6 is just equation 5 in a box) but I enjoyed it a lot nonetheless.

[ Parent | on HN ]

Re:Ghosts of Softmax: Zeros of the partition function explai... by mathvoyager on Tuesday March 17, 2026 @01:55PM (#47415990)

How does it differ from traditional "small step good, big step bad" literature?

[ Parent | on HN ]