A question about inverse problem

[Lastilla]

@lululxvi
I`m so sorry to bother you. I have some doubt about inverse problem. In PINN, we set some points as boundary condition:

    [ob_x, ob_y, ob_u, ob_v, ob_p, ob_theta] = database()
       
    ob_xy = np.hstack((ob_x, ob_y))
    
    observe_u = dde.icbc.PointSetBC(ob_xy, ob_u, component=0)
    observe_v = dde.icbc.PointSetBC(ob_xy, ob_v, component=1)

but after trainning, we need input points of the same number we input.
I input 10 points to predict, and model would output 10 points, just like I tell librarian the book index, and he give the book I need.
This question puzzles me for a long time.

[lululxvi]

I cannot understand your question.

[Lastilla]

sry for my poor interpretation.  I mean that: in prediction phase, can I rebuild the flow fluid through several points?

…

---Original--- From: "Lu ***@***.***> Date: Mon, Dec 12, 2022 04:56 AM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [lululxvi/deepxde] A question about inverse problem (Discussion#1066) I cannot understand your question. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[lululxvi]

You can predict the u/p at any point.