Does a small gradient alone establish a local minimum in a nonconvex objective?
MathematicsLevel: IntermediateMultivariate optimizer reasoning
Does a small gradient alone establish a local minimum in a nonconvex objective?
Hints
- Stationarity is necessary, not sufficient.
- Look for negative curvature.
Answer
No. Saddles and maxima can also have zero gradient. Examine curvature or nearby feasible perturbations and compare alternative starts.
Source and licence: SciPy — scalar and multivariate minimization · BSD-3-Clause · answer: Quantoros editorial