A model with two weights is being trained, and the optimizer has stopped at four different points where the loss is perfectly flat. At each one it recorded a symmetric curvature matrix , which describes how the loss bends as you move away from that point.
As Chapter 9 noted, a flat point is a true minimum, with the loss curving upward in every direction, when every eigenvalue of its curvature matrix is positive. A symmetric matrix whose eigenvalues are all strictly positive is called positive definite.
Which curvature matrix is positive definite, making its point a true minimum?
Select all that apply.