Thesis
Bi-level optimization : optimistic and pessimistic algorithm for conflicting objectives in metabolic engineering
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2024
- Thesis identifier
- T17326
- Person Identifier (Local)
- 201991298
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- In recent years, bi-level programming has emerged as a pivotal concept in the realm of optimization, addressing the inherent complexity found in hierarchical decision-making environments. This intricate framework involves two levels of decision-makers, termed the leader and the follower, each with distinct objectives and constraints that are interdependent. The leader’s decisions influence the follower’s feasible set, while the follower’s decisions, in turn, impact the leader’s outcomes. Such a dynamic interplay naturally lends itself to modelling via bi-level programming, capturing the essence of interactions as seen in Stackelberg games (Shi, G. Zhang, and Lu 2005; Stackelberg 2011; Dempe 2015). The significance of bi-level programming is further exemplified by its wide array of applications across various domains. However, the structure of bilevel problems invariably renders them NP-hard, posing significant computational challenges (Jeroslow 1985; J. F. Bard 1991). This complexity necessitates the development of efficient solution methodologies, such as reformulation techniques and sophisticated algorithmic strategies, to render these problems tractable (Deng 1998; Pineda, Bylling, and Morales 2017; Fischetti 2018). One domain where bi-level programming has been extensively applied is network optimization, exemplified by bi-level flow and routing problems. These problems often involve scenarios where a leader, such as a toll authority, seeks to optimize toll pricing strategies, while the follower, representing network users, optimizes their routing decisions based on the imposed tolls (P. Marcotte 1986; Labb´e, Patrice Marcotte, and Gilles Savard 1998a). The iterative interplay between pricing strategies and user responses in congested networks aptly demonstrates the power of bi-level models to capture and solve real-world problems with hierarchical decision structures. Another noteworthy application is found in the field of metabolic engineering, where bi level optimization frameworks are employed to redesign metabolic networks for enhanced biochemical production. In this context, the leader (engineer) aims to maximize the production of specific chemicals through genetic modifications, while the follower (microbe) naturally seeks to maximize growth, creating a complex bi-objective landscape (Burgard, Pharkya, and Maranas 2003; Tepper and Shlomi 2010). The innovative application of bi-level programming in this arena underscores its potential to drive advancements in biotechnology and industrial bioprocesses. Furthermore, the versatility of bi-level programming extends to network interdiction problems, where an attacker (leader) seeks to disrupt a network optimized by a defender (follower) to minimize losses or maximize flow (Smith, Prince, and Geunes 2013). These models are integral to understanding and mitigating vulnerabilities in critical infrastructure and defense systems. In light of the challenging nature of bi-level problems, our research focuses on developing efficient solution techniques, anchored in cutting-plane and branchand-bound methodologies. By exploring both strong duality and KKT-based reformulations, we strive to advance the state-of-the-art in solving bi-level optimization problems, offering novel insights and solutions to enhancing chemical production in metabolic networks. This dissertation contributes to the broader understanding and application of bi-level programming by elucidating its theoretical underpinnings, computational strategies, and real-world implications. As it delves deeper into solution methodologies and their applications, it aims to bridge the gap between theoretical constructs and practical implementation, Ultimately, providing a novel algorithm that will facilitate more informed decision-making processes across disciplines characterized by hierarchical interactions with special attention to metabolic engineering and chemical production.
- Advisor / supervisor
- Doostmohammadi, Mahdi
- Arulselvan, Ashwin
- Resource Type
- DOI
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of thesis T17326 | 2025-05-14 | Public | Download |