spatial_domain = dde.geometry.Polygon and self.normal = self.normal / np.linalg.norm(self.normal, axis=1).reshape(-1, 1)

[Chaztikov]

Hi, I'm wondering if there is a quick solution to the following problem:
(I apologize for my brevity, I hope I don't sound rude)

spatial_domain = dde.geometry.Polygon(pts_boundary)
yields
/home/chaztikov/.local/lib/python3.6/site-packages/deepxde/geometry/geometry_2d.py:353: RuntimeWarning: invalid value encountered in true_divide
self.normal = self.normal / np.linalg.norm(self.normal, axis=1).reshape(-1, 1)

What is the cause of this issue? Too many points? Repeated points? Corners?

If I remove the corners, is that sufficient? If I randomly sample too few points, and retry with

    nboundarypts = int(pts_boundary.shape[0] * 0.01 )
    print(nboundarypts)
    iboundarypts = np.random.choice(pts_boundary.shape[0], nboundarypts)                
    spatial_domain = dde.geometry.Polygon(pts_boundary[iboundarypts])

Then

    pts = spatial_domain.uniform_boundary_points( nboundarypts )
    plt.scatter(pts.T[0],pts.T[1]);plt.show()

Yields something like the first figure (the second figure is the original dataset)

[lululxvi]

What is the value of pts_boundary?