Abstract:We introduce two-stage stochastic min-max and min-min integer programs with bi-parameterized recourse (BTSPs), where the first-stage decisions affect both the objective function and the feasible region of the second-stage problem. To solve these programs efficiently, we introduce Lagrangian-integrated L-shaped ($L^2$) methods, which guarantee exact solutions when the first-stage decisions are pure binary. For mixed-binary first-stage programs, we present a regularization-augmented variant of this method. We also introduce distributionally robust bi-parameterized two-stage stochastic integer programs and present an extension of the $L^2$ method and a reformulation-based method for programs with finite and continuous supports, respectively. Our computational results show that the $L^2$ method surpasses the benchmark method for bi-parameterized stochastic network interdiction problems, solving all instances in 23 seconds on average, whereas the benchmark method failed to solve any instance within 3600 seconds. Additionally, it achieves optimal solutions up to 18.4 and 1.7 times faster for instances of risk-neutral and distributionally robust bi-parameterized stochastic facility location problems, respectively. Furthermore, BTSPs have applications in solving stochastic problems with decision-dependent probability distributions or sets of distributions (ambiguity set). The $L^2$ method outperforms existing approaches, achieving optimal solutions 5.3 times faster for distributionally robust facility location problem with a decision-dependent and non-relatively complete ambiguity set.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: **How to efficiently solve two - stage stochastic min - max and min - min mixed - integer programming problems (BTSPs) with bi - parameterized recourse**.
Specifically, the paper focuses on the following types of problems:
1. **Stochastic optimization problems under decision - dependent uncertainty**: In these problems, the decisions in the first stage not only affect the objective function in the second stage, but also the constraint conditions in the feasible region of the second stage. This makes the traditional two - stage stochastic programming methods no longer applicable.
2. **Distributionally robust optimization problems with decision - dependent fuzzy sets**: This type of problem takes into account the distribution fuzziness of uncertainty, and this fuzziness depends on the initial decision. Therefore, new methods need to be developed to handle this complexity.
3. **Applications in network interdiction models**: Especially in the security field, such as combating illegal supply chains, military operation planning, etc., these application scenarios require efficient algorithms to solve complex decision - making problems.
To meet the above challenges, the author introduced a new method called Lagrangian - integrated L - shaped (L2). This method can ensure an exact solution when the first - stage decision is pure binary, and also proposes a regularized enhanced variant for the first - stage program with mixed binary. In addition, the paper extends the L2 method to solve distributionally robust optimization problems with finite and continuous support, and demonstrates its superior performance on multiple instances.
### Formula summary
- **Objective function**:
\[
\min_{x\in X}\left\{f(x)+\mathbb{E}_{\tilde{\omega}}[Q(x,\tilde{\omega})]\right\}
\]
where \(Q(x,\omega)\) is defined as:
\[
Q(x,\omega):=\min/\max\left\{q(\omega)^T y + x^T G(\omega)y:W(\omega)y = r(\omega)-T(\omega)x,y\in Y\right\}
\]
- **Lagrangian dual**:
\[
D(x,\omega):=\min_{\lambda(\omega)\in\mathbb{R}^{n_1}}\left\{L(x,\lambda(\omega),\omega)-\lambda(\omega)^T x\right\}
\]
- **Improved objective function**:
\[
\min_{x\in X}\left\{f(x)-\sigma^T x+\sum_{\omega\in\Omega}p(\omega)\hat{Q}(x,\omega)\right\}
\]
where,
\[
\hat{Q}(x,\omega)=\max\left\{q(\omega)^T y + x^T G(\omega)y+\sum_{i\in[n_1]}\sigma_i(2x_i - 1)z_i:T(\omega)z+W(\omega)y = r(\omega),z\in X,y\in Y\right\}
\]
Through these formulas and methods, the paper successfully solves the non - convexity and discontinuity problems that are difficult to handle by traditional methods, thereby improving the solution efficiency and accuracy.