Timeline for What is/are the default filters used by Keras Convolution2d()?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 15, 2022 at 0:03 | comment | added | skan | Then those filters add a new set of parameters that also need to be optimized along with the weights? Does Keras treat both of these "sets of parameters" in the same way? | |
| Nov 30, 2018 at 1:04 | history | edited | timleathart | CC BY-SA 4.0 |
improved formatting
|
| Jan 23, 2017 at 22:32 | comment | added | timleathart |
@NeilSlater Thanks for your comment -- you're right, I had confused glorot_normal and glorot_uniform, and I've updated the answer to reflect this. I also added a bit of extra info about how the filters end up, as you suggested.
|
|
| Jan 23, 2017 at 22:32 | history | edited | timleathart | CC BY-SA 3.0 |
fixed normal/uniform distribution, added detail about how the filters are trained
|
| Jan 23, 2017 at 18:14 | vote | accept | ChrisFal | ||
| Jan 23, 2017 at 18:11 | comment | added | ChrisFal | Tim, thanks for providing the math. @Neil Slater - your insight that the filters, after training with backpropagation, might end up looking like edge detection, etc., was quite helpful. If I had more reputation, I would +1 both of your contributions. | |
| Jan 23, 2017 at 14:52 | comment | added | Neil Slater |
glorot_uniform does not use the normal distribution. I think you are describing glorot_normal. I don't think that matters greatly to the answer - the key points are random initialisation followed by effects of training. Might be worth explaining how the trained filters end up looking like edge/corner etc filters (maybe with one of the classic images of before/after training imaging first layer filters).
|
|
| Jan 23, 2017 at 12:08 | history | answered | timleathart | CC BY-SA 3.0 |