Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
cpldcpu
29 days ago
|
parent
|
context
|
favorite
| on:
Implementing Neural Networks on a "10-cent" RISC-V...
The latter one. The network is trained in full precision (this is required for the gradient calculation), but the weights are nudged towards the quantized values.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: