Transformer-based Stagewise Decomposition for Large-Scale Multistage Stochastic Optimization
Jul 24, 2023
Speakers
Chanyeong Kim
Speaker · 0 followers
Jongwoong Park
Speaker · 0 followers
Hyunglip Bae
Speaker · 0 followers
Woo Chang Kim
Speaker · 0 followers
About
A Transformer-based stagewise decomposition algorithm, TranSDDP, is proposed for addressing large-scale multistage stochastic programming (MSP) problems. Common stagewise decomposition algorithms employed in multistage stochastic programming, including stochastic dual dynamic programming (SDDP), approximate value functions as piecewise linear convex functions by gradually accumulating subgradient cutting planes from primal and dual solutions of stagewise subproblems. However, these methods suffer from growing time complexity as the size of the subproblems and the number of problems to solve increases. TranSDDP addresses this issue by utilizing a sequential approach for adding subgradient cutting planes to approximate the value function, leveraging the structure of the Transformer. Numerical experiments demonstrate that TranSDDP can efficiently solve MSP problems by generating a piecewise linear approximation for the value function, resulting in a significant reduction in computation time without compromising solution quality.A Transformer-based stagewise decomposition algorithm, TranSDDP, is proposed for addressing large-scale multistage stochastic programming (MSP) problems. Common stagewise decomposition algorithms employed in multistage stochastic programming, including stochastic dual dynamic programming (SDDP), approximate value functions as piecewise linear convex functions by gradually accumulating subgradient cutting planes from primal and dual solutions of stagewise subproblems. However, these methods suffe…
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Linkless Link Prediction via Relational Distillation
Watch later
Zhichun Guo, …
Conditional Graph Information Bottleneck for Molecular Relational Learning
Watch later
Grammar-Induced Geometry for Data-Efficient Molecular Property Prediction
Watch later
Minghao Guo, …
