EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
543. Tran2SP: Transformer-based Two-Stage Stochastic Programming
Invited abstract in session WA-63: Novel Optimization Models in Finance, stream OR in Banking, Finance and Insurance: New Tools for Risk Management.
Wednesday, 8:30-10:00Room: S14 (building: 101)
Authors (first author is the speaker)
1. | Woo Chang Kim
|
Industrial and Systems Engineering, KAIST |
Abstract
This talk introduces Tran2SP, a novel framework for solving two-stage stochastic programming (2SP) problems, which are characterized by making decisions under uncertainty. Traditional methods, while effective, often suffer from computational inefficiencies and the necessity to re-solve for new or altered instances. Tran2SP overcomes these challenges by employing the Transformer architecture to approximate the value function with piecewise linear functions, enhancing scalability and reducing computational demands. Unlike previous approaches that require extensive data or re-learning for modifications in the problem, Tran2SP adapts to changes efficiently, offering high-quality, robust solutions across various problem instances without the need to start from scratch. This advancement represents a significant improvement in solving 2SP problems, highlighting Tran2SP's potential to address the dynamic nature of uncertainties in optimization tasks. We further validate this approach through comprehensive experiments, confirming its effectiveness and robustness in practical applications.
Keywords
- Optimization Modeling
- Stochastic Optimization
- Machine Learning
Status: accepted
Back to the list of papers