A hybrid evolutionary algorithm for influence maximization in complex networks using invasive weed optimization and...

Area of Science:

• Computer Science
• Artificial Intelligence
• Optimization Algorithms

Background:

• Social networks are vital for information spread, but optimizing their vast search spaces is difficult.
• Existing optimization methods struggle with the rugged landscapes of large-scale problems.

Purpose of the Study:

• To introduce a novel Hybrid Weed-Gravitational Evolutionary Algorithm (HWGEA) for robust continuous optimization.
• To develop a discrete variant (DHWGEA) for efficient influence maximization in social networks.
• To enhance the balance between exploration and exploitation in optimization.

Main Methods:

• HWGEA unifies Invasive Weed Optimization and Gravitational Search with adaptive mutation.
• DHWGEA incorporates topology-aware initialization, dynamic local search, and an Expected Influence Score surrogate.
• Algorithms were tested on 23 continuous benchmarks and real-world engineering designs.

Main Results:

• HWGEA achieved superior performance on continuous benchmarks, outperforming several established algorithms.
• HWGEA demonstrated competitive results on engineering design problems.
• DHWGEA provided a practical accuracy-efficiency trade-off for influence maximization, outperforming PageRank.

Conclusions:

• HWGEA and DHWGEA offer a cohesive, scalable framework for both continuous and discrete optimization.
• The adaptive components of the algorithms reduce the need for manual parameter tuning.
• These algorithms provide a robust and efficient approach to complex optimization challenges in computer science and AI.