On the Rate of Convergence of Optimal Solutions of Monte Carlo Approximations of Stochastic Programs
read more
Citations
Lectures on Stochastic Programming: Modeling and Theory
Distributionally Robust Optimization Under Moment Uncertainty with Application to Data-Driven Problems
Monte Carlo Sampling Methods
Uncertain convex programs: randomized solutions and confidence levels
A Sample Approximation Approach for Optimization with Probabilistic Constraints
References
Large Deviations Techniques and Applications
Convex analysis and minimization algorithms
Perturbation Analysis of Optimization Problems
Related Papers (5)
Frequently Asked Questions (9)
Q2. What is the optimal solution of the problem?
The authors assume that the distribution of Y is symmetrical around zero and hence x is an optimal solution of the true problem Suppose now that the distribution of Y is continuous with density function g y
Q3. what is the simplest way to show that f x is a positive homo gene?
Note that since f x is a positively homo geneous convex real valued and hence continuous function it follows from that f x d kdk for some and all d T x
Q4. what is the second proof of AN?
Let us nally observe that since f x and hence fN x are linear on A and AN is the set of minimizers of fN x over A it follows that AN is a face of ALet us give now the second proof
Q5. What is the true problem with the IRm?
It follows from the assumptions i and ii that the expected value function f x is piecewise linear and convex and hence f x can be represented as amaximum of a nite number of a ne functions i x i n Consequently the space IRm can be partitioned into a union of convex polyhedral sets C
Q6. what is the AN of the approximating problem?
This shows that w p for N large enough the set AN of optimal solutions of the approximating problem is non emptySince f x is piecewise linear and convex the authors have that subdi erentials of f x are convex compact polyhedral sets and by Lemma it follows that the total number of the extreme points of all subdi erentials f x is nite Moreover since for any x A the authors have that f x it follows that there exists such that the distance from the null vector IRm to f x is greater than for all x A Together withthis implies that w p for N large enough fN x for all x A and hence any x A cannot be an optimal solution of the approximating problem
Q7. What is the optimal solution of the approximating problem?
By the de nition of the set F the authors have that if N F then N d for all d T x Sm Consequently in that case xN x is the unique optimal solution of the approximating problem
Q8. what is the rst order i e linear growth of f x?
Under such conditions rst order i e linear growth of f x holds globally i e for all xTheorem Suppose that i for every the function h is convex ii the expected value function f is well de ned and is nite valued iii the set is closed and convex iv assumption A holds
Q9. what is the AN of optimal solutions of the approximating problem?
Therefore w p for N large enough the authors have that fN xi f xi for i f g and fN xj f xj for j f qg and hence condition follows Together with assertion c of Lemma this proves that AN is non empty and forms a face of AUnder the assumptions of the above theorem the set AN of optimal solutions of the approximating problem is convex and polyhedral