Efficient Parallel Genetic Algorithm for Perturbed Substructure Optimization in Complex Network (2412.20980v1)

Abstract: Evolutionary computing, particularly genetic algorithm (GA), is a combinatorial optimization method inspired by natural selection and the transmission of genetic information, which is widely used to identify optimal solutions to complex problems through simulated programming and iteration. Due to its strong adaptability, flexibility, and robustness, GA has shown significant performance and potentiality on perturbed substructure optimization (PSSO), an important graph mining problem that achieves its goals by modifying network structures. However, the efficiency and practicality of GA-based PSSO face enormous challenges due to the complexity and diversity of application scenarios. While some research has explored acceleration frameworks in evolutionary computing, their performance on PSSO remains limited due to a lack of scenario generalizability. Based on these, this paper is the first to present the GA-based PSSO Acceleration framework (GAPA), which simplifies the GA development process and supports distributed acceleration. Specifically, it reconstructs the genetic operation and designs a development framework for efficient parallel acceleration. Meanwhile, GAPA includes an extensible library that optimizes and accelerates 10 PSSO algorithms, covering 4 crucial tasks for graph mining. Comprehensive experiments on 18 datasets across 4 tasks and 10 algorithms effectively demonstrate the superiority of GAPA, achieving an average of 4x the acceleration of Evox. The repository is in https://github.com/NetAlsGroup/GAPA.

• The paper introduces GAPA, a framework that accelerates genetic algorithms for perturbed substructure optimization in complex networks using GPU parallelism and optimized matrix operations.
• GAPA improves efficiency by restructuring genetic operations and fitness functions into matrix form, demonstrating up to fourfold acceleration over existing methods across various network tasks.
• This framework offers a practical approach to leverage parallel computing for faster, robust network optimization, applicable in areas like social, biological, and infrastructure network analysis.

Efficient Parallel Genetic Algorithm for Perturbed Substructure Optimization in Complex Networks

The paper details the development and implementation of GAPA, an innovative framework designed explicitly to accelerate Genetic Algorithms (GAs) for solving Perturbed Substructure Optimization (PSSO) problems in complex networks. The framework leverages evolutionary computing principles and GPU-based parallel acceleration to address the computational challenges inherent in GA-based PSSO problems.

GAPA is motivated by the need to enhance the efficiency of GAs in identifying optimal solutions for PSSO tasks, which involve finding optimal perturbation structures within large and complex network topologies. The framework aims to mitigate the challenges posed by the intricacies and diversity of PSSO applications, particularly when dealing with large-scale data.

Key Contributions

1. Genetic Operations Optimization: The framework simplifies traditional genetic operations—including population initialization, crossover, mutation, fitness calculation, and elitism—by restructuring them into matrix operations. This transformation enables parallel processing, making it particularly well-suited for GPU acceleration. The paper demonstrates that these optimizations can reduce the GA's computation time significantly, thereby enhancing the efficiency of solving PSSO problems.
2. Fitness Function Design: Recognizing the complexity of fitness evaluations in PSSO tasks, the authors propose a strategy to design fitness functions that facilitate efficient parallel computation. An illustrative case based on the Six Degrees of Separation Theory is presented, showcasing a significant reduction in the time complexity for computing connectivity metrics in network graphs using matrix power calculations.
3. Acceleration Modes: The framework introduces different acceleration modes, ranging from single-process to distributed multi-process environments, to maximize computational efficiency. These modes allow users to adapt the framework according to the specific computational resources available, thereby ensuring broad applicability across diverse environments.
4. Comprehensive Experimentation: The authors conduct experiments across 18 datasets spanning four core graph mining tasks—community detection attack, critical node detection, node classification attack, and link prediction attack—demonstrating GAPA's ability to achieve up to fourfold acceleration over existing methods like Evox. This showcases GAPA's effectiveness in improving GA execution efficiency while maintaining the quality of solutions.

Implications and Future Directions

The paper highlights the significant computational benefits of leveraging parallel computing architectures for evolutionary algorithms in complex network environments. The introduction of GAPA underscores the potential for specialized frameworks to address the practical limitations of traditional genetic algorithms.

The results indicate that GAPA could be a significant enabler for real-world applications requiring rapid and robust optimization solutions, particularly in domains such as social network analysis, biological network modeling, and infrastructure resilience assessment. The framework's ability to adapt to varying sizes and complexities of data ensures its utility in expansive applications.

Future research could explore adaptive mechanisms within GAPA to optimize parallel resource utilization dynamically and further simplify the deployment process for users. Additionally, expanding the framework to support a broader range of GA variants and incorporating advancements in hardware acceleration technologies could further enhance its applicability and performance.

Overall, this paper presents a notable advancement in the application of genetic algorithms to network optimization problems, offering both theoretical insights and practical tools for harnessing parallel acceleration to tackle the growing complexities of modern networked systems.