- Scheduling and Optimization Algorithms
- Advanced Data Storage Technologies
- Constraint Satisfaction and Optimization
- Advanced Multi-Objective Optimization Algorithms
- Vehicle Routing Optimization Methods
- Formal Methods in Verification
- Machine Learning and Data Classification
- Metaheuristic Optimization Algorithms Research
- Cloud Computing and Resource Management
- Caching and Content Delivery
- Process Optimization and Integration
- Software Testing and Debugging Techniques
- Scheduling and Timetabling Solutions
- Parallel Computing and Optimization Techniques
- Supply Chain and Inventory Management
- VLSI and FPGA Design Techniques
- Data Stream Mining Techniques
- Embedded Systems Design Techniques
- Natural Language Processing Techniques
- Rough Sets and Fuzzy Logic
- Distributed and Parallel Computing Systems
- Software Engineering Research
- Fuzzy Logic and Control Systems
- Evolutionary Algorithms and Applications
- Reinforcement Learning in Robotics
Huawei Technologies (China)
2021-2025
Huawei Technologies (United Kingdom)
2024
Huawei Technologies (Sweden)
2022-2023
University of Science and Technology of China
2021-2023
Xiangtan University
2023
Huawei Technologies (France)
2022
Xinjiang Academy of Agricultural Sciences
2022
Wuhan University
2009
Air Force General Hospital PLA
2009
Inner Mongolia University for Nationalities
2006
This work is geared toward a real-world manufacturing planning (MP) task, whose two objectives are to maximize the order fulfillment rate and minimize total cost. More important, requirements constraints in real make MP task very challenging several aspects. For example, needs cover many production components of multiple plants over 30-day horizon, which means that it involves large number decision variables. Furthermore, task's have extremely different magnitudes, some difficult handle....
This paper reviews the recent literature on solving Boolean satisfiability problem (SAT), an archetypal $$\cal{N}\cal{P}$$ -complete problem, with aid of machine learning (ML) techniques. Over last decade, society advances rapidly and surpasses human performance several tasks. trend also inspires a number works that apply methods for SAT solving. In this survey, we examine evolving ML solvers from naive classifiers handcrafted features to emerging end-to-end solvers, as well progress...
The Earth observation satellites (EOSs) scheduling problem is generally considered as a complex combinatorial optimization due to various technical constraints. It significant develop efficient computational frameworks solve this problem. In paper, an intelligent EOSs framework developed using imitation learning based on mixed integer linear programming (MILP). composed of two processes: pre-processing, modeling and solving process. the pre-processing process, analytical method generate...
Scheduling job flows efficiently and rapidly on distributed computing clusters is one of huge challenges for daily operation data centers. In a practical scenario, single consists numerous stages with complex dependency relation represented as Directed Acyclic Graph (DAG) structure. Nowa-days center usually equips cluster heterogeneous servers which are different in the hardware/software configuration. From both cost saving environmental friendliness, centers could benefit lot from...
Cutting planes (cuts) are important for solving mixed-integer linear programs (MILPs), which formulate a wide range of real-world applications. Cut selection -- aims to select proper subset the candidate cuts improve efficiency MILPs heavily depends on (P1) should be preferred, and (P2) how many selected. Although modern MILP solvers tackle (P1)-(P2) by manually designed heuristics, machine learning offers promising approach learn more effective heuristics from collected specific However,...
As one of the most critical components in modern LP solvers, presolve linear programming (LP) employs a rich set presolvers to remove different types redundancy input problems by equivalent transformations. We found from extensive experiments that routine-that is, method determining (P1) which select, (P2) what order execute, and (P3) when stop-significantly impacts efficiency solving LPs. However, designing high-quality routines is highly challenging due enormous search space, further...
In the field of many-objective evolutionary optimization algorithms (MaOEAs), how to maintain balance between convergence and diversity has been a significant research problem. With increase number objectives, mutually nondominated solutions increases rapidly, multi-objective algorithms, based on Pareto-dominated relations, become invalid because loss selection pressure in environmental selection. order solve this problem, indicator-based have proposed; however, they are not good enough at...
Industrial SAT formula generation is a critical yet challenging task. Existing approaches can hardly simultaneously capture the global structural properties and maintain plausible computational hardness. We first present an in-depth analysis for limitation of previous learning methods in reproducing hardness original instances, which may stem from inherent homogeneity their adopted split-merge procedure. On top observations that industrial formulae exhibit clear community structure oversplit...
Most combinatorial optimization problems can be formulated as mixed integer linear programming (MILP), in which branch-and-bound (B\&B) is a general and widely used method. Recently, learning to branch has become hot research topic the intersection of machine optimization. In this paper, we propose novel reinforcement learning-based B\&B algorithm. Similar offline learning, initially train on demonstration data accelerate massively. With improvement training effect, agent starts interact...
Data prefetching is important for storage system optimization and access performance improvement. Traditional prefetchers work well mining patterns of sequential logical block address (LBA) but cannot handle complex non-sequential that commonly exist in real-world applications. The state-of-the-art (SOTA) learning-based cover more LBA accesses. However, they do not adequately consider the spatial interdependencies between deltas, which leads to limited robustness. This paper proposes a novel...
An adaptive band selection algorithm for dimension reduction of hyperspectral images is proposed. Considering the spatial correlation and spectral correlation, a rule, referring to information its constructed selection. To test efficiency this algorithm, K-means unsupervised classification was applied on generated from algorithm. The results showed that proposed reduced computation amount improved accuracy.
In the past few years, there has been an explosive surge in use of machine learning (ML) techniques to address combinatorial optimization (CO) problems, especially mixed-integer linear programs (MILPs). Despite achievements, limited availability real-world instances often leads sub-optimal decisions and biased solver assessments, which motivates a suite synthetic MILP instance generation techniques. However, existing methods either rely heavily on expert-designed formulations or struggle...
Large-scale LP problems from industry usually contain much redundancy that severely hurts the efficiency and reliability of solving LPs, making presolve (i.e., problem simplification module) one most critical components in modern solvers. However, how to design high-quality routines -- is, program determining (P1) which presolvers select, (P2) what order execute, (P3) when stop remains a highly challenging task due extensive requirements on expert knowledge large search space. Due sequential...
Abstract This paper studies a real-world manufacturing problem, which is modeled as bi-objective integer programming problem. The variables and constraints involved are usually numerous dramatically vary according to the data. It very challenging directly solve such large-scale problems using heuristic algorithms or commercial solvers. Considering that decision space of sparse has block-like structure, we propose use decomposition methods accelerate optimization process. However, existing...
The Dynamic Pickup and Delivery Problem (DPDP) is an essential problem within the logistics domain. So far, research on this has mainly focused using artificial data which fails to reflect complexity of real-world problems. In draft, we would like introduce a new benchmark from real business scenarios as well simulator supporting dynamic evaluation. have been published successfully supported ICAPS 2021 competition participated by 152 teams.
Cache plays an important role to maintain high and stable performance (i.e. throughput, low tail latency throughput jitter) in storage systems. Existing rule-based cache management methods, coupled with engineers' manual configurations, cannot meet ever-growing requirements of both time-varying workloads complex systems, leading frequent overloading.
Nowadays, the Hierarchical Storage System (HSS) is considered as an ideal model to meet cost-performance demand. The data migration between storing tiers of HSS way achieve goal. bandwidth control limit maximum amount migration. Most previous research about focus on studying policy instead control. However, recent cache and networking optimization suggest that has significant impact system performance. Few work achieves a satisfactory in since it hard for so many tasks simultaneously. In...
Logic Synthesis (LS) plays a vital role in chip design -- cornerstone of the semiconductor industry. A key task LS is to transform circuits modeled by directed acyclic graphs (DAGs) into simplified with equivalent functionalities. To tackle this task, many operators apply transformations subgraphs rooted at each node on an input DAG sequentially. However, we found that large number are ineffective, which makes applying these highly time-consuming. In particular, notice runtime Resub and Mfs2...
Multiple patterning lithography (MPL) has been introduced in the integrated circuits manufacturing industry to enhance feature density as technology node advances. A crucial step of MPL is assigning layout features different masks, namely decomposition. Exact algorithms like integer linear programming (ILP) can solve decomposition optimality but lacks scalability for very dense patterns. Approximation (e.g., programming, semi-definite programming) and heuristics Exact-Cover) are capable...
Machine learning has been successfully applied to improve the efficiency of Mixed-Integer Linear Programming (MILP) solvers. However, learning-based solvers often suffer from severe performance degradation on unseen MILP instances -- especially large-scale a perturbed environment due limited diversity training distributions. To tackle this problem, we propose novel approach, which is called Adversarial Instance Augmentation and does not require know problem type for new instance generation,...
In an era of digital ubiquity, efficient resource management and decision-making are paramount across numerous industries. To this end, we present a comprehensive study on the integration machine learning (ML) techniques into Huawei Cloud's OptVerse AI Solver, which aims to mitigate scarcity real-world mathematical programming instances, surpass capabilities traditional optimization techniques. We showcase our methods for generating complex SAT MILP instances utilizing generative models that...
Cutting planes (cuts) play an important role in solving mixed-integer linear programs (MILPs), which formulate many real-world applications. Cut selection heavily depends on (P1) cuts to prefer and (P2) how select. Although modern MILP solvers tackle (P1)-(P2) by human-designed heuristics, machine learning carries the potential learn more effective heuristics. However, existing learning-based methods prefer, neglecting importance of Moreover, we observe that (P3) what order selected...