Nguyen T. V. Hang,Ebrahim Sarabi
Abstract:$\mathcal C^2$-partial smoothness of functions has been an important subject of research in optimization, on both theoretical and algorithmic aspects, since it was first introduced by Lewis in 2002. Our work aims at providing a fresh variational analysis viewpoint on the class of $\mathcal C^2$-partly smooth functions. Namely, we explore the relationship between $\mathcal C^2$-partial smoothness and strict twice epi-differentiability and demonstrate that functions from the latter class are always strictly twice epi-differentiable. On the other hand, we provide two examples to show that the opposite conclusion does not hold in general. As a consequence of our analysis, we calculate the second subderivative of $\mathcal C^2$-partly smooth functions. Applications to stability analysis of related generalized equations involving a general perturbation and to asymptotic analysis of the well-known sample average approximation method for stochastic programs with $\mathcal C^2$-partly smooth regularizers are also given.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to conduct variational analysis on \(C^2\)-partially smooth functions, especially to explore the relationship between them and strict twice epi - differentiability. Specifically:
1. **Research Background and Motivation**:
- \(C^2\)-partially smooth functions play an important role in optimization theory because they can exhibit smooth properties on certain smooth manifolds while allowing the function to remain non - smooth elsewhere. This characteristic makes such functions very useful in optimization and variational analysis.
- Strict twice epi - differentiability is an important concept for measuring the second - order variational properties of a function, and it is crucial for understanding the local behavior of the function and the convergence of optimization algorithms.
2. **Main Objectives**:
- The paper aims to re - examine \(C^2\)-partially smooth functions from the perspective of variational analysis and explore whether these functions are always strictly twice epi - differentiable.
- Through rigorous mathematical proofs, the author shows that functions belonging to the \(C^2\)-partially smooth class are indeed always strictly twice epi - differentiable.
- At the same time, the author provides two counter - examples to illustrate that the opposite conclusion does not always hold, that is, a strictly twice epi - differentiable function does not necessarily belong to the \(C^2\)-partially smooth class.
3. **Applications and Significance**:
- The research results help to further understand the second - order variational properties of \(C^2\)-partially smooth functions, thereby providing theoretical support for the stability and asymptotic analysis of related optimization problems.
- Specific applications include the stability analysis of generalized equations and the asymptotic behavior analysis of the sample average approximation method in stochastic programming.
### Key Formulas
- **Definition of Strict Twice Epi - Differentiability**:
A function \( f \) is strictly twice epi - differentiable at a point \( \bar{x} \) with respect to \( \bar{v}\in\partial f(\bar{x}) \) if there exists a function \( d^2f(\bar{x}, \bar{v}) \) such that the parameterized second - order difference quotient function \( \Delta_t^2 f(\bar{x}, \bar{v})(w) \) epi - converges to \( d^2f(\bar{x}, \bar{v})(w) \) as \( t\rightarrow0 \) and \( (x, v)\rightarrow(\bar{x}, \bar{v}) \).
\[
d^2f(\bar{x}, \bar{v})(w)=\liminf_{t\searrow0, w'\to w}\frac{f(\bar{x}+tw') - f(\bar{x})-t\langle\bar{v}, w'\rangle}{\frac{1}{2}t^2}
\]
- **Calculation of Second - Order Sub - derivatives**:
For \( C^2 \)-partially smooth functions, the paper derives the specific expressions of their second - order sub - derivatives, which provides a basis for further theoretical analysis and applications.
### Conclusion
Through the above research, the paper not only deepens the understanding of \( C^2 \)-partially smooth functions but also provides new tools and perspectives for stability analysis and algorithm design in optimization theory.