EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1883. Recent Advances in Solving Large-Scale Optimizations from an Industry Perspective
Invited abstract in session WA-29: Advances in Optimization for Industrial Applications, stream Combinatorial Optimization.
Wednesday, 8:30-10:00Room: 157 (building: 208)
Authors (first author is the speaker)
| 1. | Fan Zhang
|
| Huawei Hong Kong Research Centre | |
| 2. | Yingying Cheng
|
| Huawei Hong Kong Research Centre | |
| 3. | Jiazheng Wang
|
| Huawei Hong Kong Research Centre | |
| 4. | Junyan Liu
|
| Huawei Hong Kong Research Center | |
| 5. | Yongfeng Li
|
| Huawei Technologies Co., Ltd., Theory Lab | |
| 6. | JIE SUN
|
| Huawei Hong Kong Research Center | |
| 7. | Jianshu Li
|
| 8. | Kun Mao
|
| , |
Abstract
In this talk, we introduce large-scale mathematical optimization problems that arise within the complex landscape of the information and communication industry. Our focus spans several practical scenarios, including resource allocation in large scale data networks, traffic planning in optical networks, as well as task allocation and scheduling in deployment operations. These optimization challenges predominantly fall into the categories of LP or MILP. The scale of these problems varies significantly, ranging from millions to billions of variables and constraints. Some medium-sized problems necessitate real-time decision-making, with solving times in milliseconds to seconds. The larger problems demand hours of computational effort for offline planning. The stringent requirement poses considerable challenges, especially given the constraints of single-server computational resources. We aim to strike a balance between customized and generic design approaches across multiple dimensions, including optimality, efficiency, robustness, and generality. Leveraging Huawei’s OptVerse optimization solver, the design solution directly translates to increased revenue or reduced costs. Additionally, some contributions further enhance OptVerse’s generic problem-solving performance. Lastly, we explore the concept of co-optimization of software and hardware—a new dimension that promises innovative solution design, novel development modes, and potential business models for future solvers.
Keywords
- Large Scale Optimization
- Engineering Optimization
- Mathematical Programming
Status: accepted
Back to the list of papers