G. C. Bento,J. X. da Cruz Neto
Abstract:In this paper we proved that the sequence generated by the proximal point method, associated to a unconstrained optimization problem in the Riemannian context, has finite termination when the objective function has a weak sharp minima on the solution set of the problem.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: **Prove the finite termination of the Proximal Point Method on Hadamard manifolds when the objective function has weakly sharp minima on the solution set**.
Specifically, the paper studies the following optimization problem:
\[
\text{(P)} \quad \min_{p \in M} f(p)
\]
where \( M \) is a complete Riemannian manifold and \( f: M \to \mathbb{R} \) is a convex function. For the initial point \( p_0 \in M \), the Proximal Point Method generates a sequence \( \{p_k\} \subset M \) as follows:
\[
p_{k + 1} \in \arg\min_{p \in M} \left\{ f(p)+\lambda_k d^2(p, p_k) \right\}
\]
where \( \{\lambda_k\} \) is a sequence of positive numbers and \( d \) is the Riemannian distance.
The main contribution of the paper lies in proving that when the objective function \( f \) has weakly sharp minima on the solution set, the Proximal Point Method terminates in a finite number of steps. The concept of weakly sharp minima was extended by Burke and Ferris to include cases with multiple solutions and has been further studied on Riemannian manifolds.
### Key Concepts
1. **Weak Sharp Minima**:
Let \( U \) be the solution set of problem (P). If there exists \( \alpha>0 \) such that for any \( p \in U \) and \( q \in M \), we have
\[
f(q) \geq f(p)+\alpha d_U(q)
\]
where \( d_U(q)=\inf_{p \in U} d(q, p) \), then \( U \) is called the weakly sharp minima set of problem (P).
2. **Hadamard Manifold**:
A Hadamard manifold is a Riemannian manifold that is complete, simply - connected, and has non - positive sectional curvature. Such manifolds have geometric properties similar to Euclidean spaces, for example, there exists a unique shortest geodesic between any two points.
3. **Proximal Point Method**:
The Proximal Point Method is an iterative algorithm for solving optimization problems. It ensures that the solution at each step is closer to the optimal solution by introducing an additional distance term.
### Main Results
The core result of the paper is Theorem 3.1, which shows that if the solution set \( U \) is the weakly sharp minima set of problem (P), then the Proximal Point Method terminates in a finite number of steps. This result depends on the following key propositions:
- **Proposition 3.1**: If \( M \) is a Hadamard manifold, then the sequence \( \{p_k\} \) generated by the Proximal Point Method is well - defined and satisfies
\[
\lambda_k (\exp^{-1}_{p_{k + 1}} p_k) \in \partial f(p_{k + 1})
\]
- **Proposition 3.2**: A necessary and sufficient condition for the solution set \( U \) to be a weakly sharp minima set is
\[
\alpha B_p \cap N_U(p) \subset \partial f(p), \quad p \in U
\]
- **Proposition 3.3**: If \( U \) is a weakly sharp minima set and there exist \( q \in M \) and \( w \in T_q M \) such that \( \|w\| < \alpha \) and \( w \in \partial f(q) \), then \( q \in U \).
These results together prove that the Proximal Point Method has finite termination under the weakly sharp minima condition, thus providing a theoretical basis for solving optimization problems.