Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
yobbo
on April 30, 2022
|
parent
|
context
|
favorite
| on:
Why train when you can optimize?
It can, but sin(x) has infinite number of extremes, and the gradients will vanish at those points. Activations will get stuck at 1 and -1 (x=π/2, 3π/2, ...). They set x+(1/a)*sin²(x) to be monotonic, which fixes this.
Or you need to optimize without using gradients.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
Or you need to optimize without using gradients.