Bregman Finito/MISO for Nonconvex Regularized Finite Sum Minimization without Lipschitz Gradient Continuity.
Puya LatafatAndreas ThemelisMasoud AhookhoshPanagiotis PatrinosPublished in: SIAM J. Optim. (2022)
Keyphrases
- objective function
- total variation
- half quadratic
- convex optimization
- weighted sum
- norm minimization
- optimization problems
- bregman divergences
- pointwise
- convex functions
- lower bound
- regularization term
- image restoration and reconstruction
- finite number
- least squares
- cost function
- maximum entropy
- linear programming
- quasi newton method
- global optimization
- linear program
- gradient information
- stationary points
- risk minimization
- image restoration
- metric space
- edge preserving
- loss function
- absolute difference
- edge detection