Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
jwuphysics
7 months ago
|
parent
|
context
|
favorite
| on:
Bayesian Neural Networks
Could you say a bit more about how so?
pkoird
7 months ago
[–]
KANs have learnable activations based on splines parameterized on few variables. You can specify a prior over those variables, effectively establishing a prior over your activation function.
Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
