RegTR
RegTR copied to clipboard
yewzijian
A different RR appears for each test
Why the RR results are different every time when using the same model for testing?
The neighbourhood computation during preprocessing gives slightly different points each time, so the result will differ a bit.
On Fri, 2 Jun 2023 at 11:41 PM, deasa99 @.***> wrote:
Why the RR results are different every time when using the same model for testing?
— Reply to this email directly, view it on GitHub https://github.com/yewzijian/RegTR/issues/18, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADIBP6622I7EM7OAHSDKIQLXJICTZANCNFSM6AAAAAAYYPMYPA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hello, thanks for your answer! So this is a normal phenomenon, is there any way to make it not change?
Unfortunately not easily, without replacing the preprocessing code. However, you can replace the codes with the CPU-based ones which will give deterministic results, at the cost of speed.
On Sat, 3 Jun 2023 at 9:33 AM, deasa99 @.***> wrote:
Hello, thanks for your answer! So this is a normal phenomenon, is there any way to make it not change?
— Reply to this email directly, view it on GitHub https://github.com/yewzijian/RegTR/issues/18#issuecomment-1574538609, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADIBP67QYAI7TQHS7ZF6QPLXJKIABANCNFSM6AAAAAAYYPMYPA . You are receiving this because you commented.Message ID: @.***>
Yes, that’s the expected behaviour.
On Sat, 3 Jun 2023 at 10:19 AM, deasa99 @.***> wrote:
Thank you for your answer! Can this be understood as a normal phenomenon?
— Reply to this email directly, view it on GitHub https://github.com/yewzijian/RegTR/issues/18#issuecomment-1574571986, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADIBP65AL3JUEYWA5SEU563XJKNJ7ANCNFSM6AAAAAAYYPMYPA . You are receiving this because you commented.Message ID: @.***>