Why the hessian needs to be positive definite to distinguish extrema?

194 Views Asked by At

I have a test in 2 days and I can't figure out why I need the hessian to be positive definite to distinguish extrema. In other words, why do we need the condition $$w^{T} \cdot \nabla_{x}^{2}L(a,\lambda) \cdot w > 0$$

1

There are 1 best solutions below

2
On

Hint: Write the Taylor expansion up to second derivative .