Timeline for Why does a zero-input network initialize all bias terms to 0? (pytorch)
Current License: CC BY-SA 4.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 19, 2022 at 8:05 | vote | accept | Mondo Duke | ||
| Oct 15, 2022 at 15:59 | answer | added | PeaBrane | timeline score: 2 | |
| Oct 14, 2022 at 7:46 | comment | added | maxy |
I don't understand why you expect only one parameter in the first case with bias=True. You have activation = weight * input + bias. One input, one output, everything is scalar. The layer has two parameters: the weight, and the bias.
|
|
| Oct 14, 2022 at 1:03 | history | asked | Mondo Duke | CC BY-SA 4.0 |