search

tiwari-research-group / Koopman-with-KANs

4 stars 2 forks source link

CAN NOT reproduce the result of Fig. 3 #1

Closed jinaoao closed 3 months ago

jinaoao commented 3 months ago

The idea that uses KANs to replace MLP in Deep Koopman Theory is interesting! However, I can not reproduce the LQR response result that you presented in the paper.

tiwari-research-group commented 3 months ago

Did it work?

Madhur Tiwari

On Thu, Jun 20, 2024 at 3:23 PM Ao Jin @.***> wrote:

Closed #1 https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1 as completed.

— Reply to this email directly, view it on GitHub https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1#event-13226561849, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALK5JYOLBCK663LTHK3H6ZLZIKRCRAVCNFSM6AAAAABJTQ4W22VHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJTGIZDMNJWGE4DIOI . You are receiving this because you are subscribed to this thread.Message ID: <tiwari-research-group/Koopman-with-KANs/issue/1/issue_event/13226561849@ github.com>

jinaoao commented 3 months ago

Did it work? Madhur Tiwari … On Thu, Jun 20, 2024 at 3:23 PM Ao Jin @.***> wrote: Closed #1 <#1> as completed. — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALK5JYOLBCK663LTHK3H6ZLZIKRCRAVCNFSM6AAAAABJTQ4W22VHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJTGIZDMNJWGE4DIOI . You are receiving this because you are subscribed to this thread.Message ID: <tiwari-research-group/Koopman-with-KANs/issue/1/issue_event/13226561849@ github.com>

No, the trajectory of pemdulum would diverge. But the system \dot{x}=(A-B*K)x is stable. So it's weird. I have another question. The control signal computed from the lifting system should be applied to the true system (nonlinear dynamics), however, the code you've presented applied the control signal to lifting system directly. Is that inappropriate?

tiwari-research-group commented 3 months ago

Let me ask my student.

Madhur Tiwari

On Fri, Jun 21, 2024 at 6:37 AM Ao Jin @.***> wrote:

Did it work? Madhur Tiwari … <#m2492128261476644933> On Thu, Jun 20, 2024 at 3:23 PM Ao Jin @.***> wrote: Closed #1 https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1 <#1 https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1> as completed. — Reply to this email directly, view it on GitHub <#1 (comment) https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1#event-13226561849>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALK5JYOLBCK663LTHK3H6ZLZIKRCRAVCNFSM6AAAAABJTQ4W22VHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJTGIZDMNJWGE4DIOI . You are receiving this because you are subscribed to this thread.Message ID: <tiwari-research-group/Koopman-with-KANs/issue/1/issue_event/13226561849@ github.com>

No, the trajectory of pemdulum would diverge. But the system \dot{x}=(A-B*K)x is stable. So it's weird. I have another question. The control signal computed from the lifted system should be applied to the true system (nonlinear dynamics), however, the code you've presented applied the control signal to lifting system directly. Is that inappropriate?

— Reply to this email directly, view it on GitHub https://github.com/tiwari-research-group/Koopman-with-KANs/issues/1#issuecomment-2181802931, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALK5JYOOVKNEXW6XR76TQLLZIN4DXAVCNFSM6AAAAABJTQ4W22VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOBRHAYDEOJTGE . You are receiving this because you commented.Message ID: @.***>