Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
antupis
on Jan 8, 2020
|
parent
|
context
|
favorite
| on:
Training GPT-2 to Play Chess
changing softmax to something else might fix that when there is a limited number of good moves softmax is far from optimal.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: