GNU Astronomy Utilities - Tasks: task #16362, Library: The pooling program for...
You are not allowed to post comments on this tracker with your current authentication level.
task #16362: Library: The pooling program for the Gnuastro
Submitter: | Faezeh Bidjarchian <fbidjarchian> | ||
Submitted: | Thu 04 May 2023 08:43:58 PM UTC | ||
Should Start On: | Thu 04 May 2023 12:00:00 AM UTC | Should be Finished on: | Thu 04 May 2023 12:00:00 AM UTC |
Category: | Libraries | Priority: | 5 - Normal |
Item Group: | Enhancement | Status: | None |
Privacy: | Public | Assigned to: | None |
Percent Complete: | 80% | Open/Closed: | Open |
Effort: | 0.00 |
Sat 08 Jul 2023 03:02:29 PM UTC, comment #3: |
Faezeh Bidjarchian <fbidjarchian> |
Fri 05 May 2023 12:50:07 PM UTC, comment #2: Many thanks for your great comments.
|
Faezeh Bidjarchian <fbidjarchian> |
Fri 05 May 2023 11:44:33 AM UTC, comment #1: Thanks Faezeh! This is a useful operator in many scenarios ;-).
|
Mohammad Akhlaghi <makhlaghi> |
Thu 04 May 2023 08:43:58 PM UTC, original submission:
In machine learning, pooling is a technique to decrease the complexity of the images (feature maps).
That image is your desired input, 2 is the size of pooling, and "pool-max" is the selected operator.
|
Faezeh Bidjarchian <fbidjarchian> |
No files currently attached
Depends on the following items: None found
Items that depend on this one: None found
There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.
Follow 3 latest changes.
Date | Changed by | Updated Field | Previous Value | => | Replaced by |
---|---|---|---|---|---|
2023-05-05 | makhlaghi | Category | None | Libraries | |
Item Group | None | Enhancement | |||
Percent Complete | 0% | 80% |
Powered by Savane 3.13-02a9.
Corresponding source code
We are expecting the selected pixels to be located in the center of the pooling window.
Because the pooling is a general operator and we want to use it in another application like WCS. For example, we don't want to shift the input by one pixel in the sky when we select stride 1. So we must consider (like the conv layer when using "same" padding) the row and column beside the input. But this is not what the pooling operator is doing in ML.
When the CNN in Gnuastro is completed, It would be better to define such a function and check how the CNN outputs change.