Expert Commentary:
This article presents an interesting research study that explores the use of genetic programming and symbolic regression to model and understand complex network structures. The authors acknowledge the growing interest in studying complex systems using network models and highlight the importance of developing generative processes to explain these networks.
The use of genetic programming and symbolic regression in this context is particularly noteworthy. By evolving computer programs that effectively explore a multidimensional search space, these techniques can iteratively find better solutions that explain network structures. The advantage of using symbolic regression is that it replicates network morphologies using both structure and processes, without relying on the scientist’s intuition or expertise. This eliminates potential biases and allows for the discovery of unbiased, interpretable rules for a range of empirical networks.
In this study, the authors extend the approach by incorporating time-varying networks. They introduce a modified generator semantics that can create and retrieve rules for networks that evolve over time. This is an important addition, as it enables the study of network dynamics and the identification of growth processes in multiple stages.
To improve the framework, the authors incorporate methods from the genetic programming toolkit, such as recombination, which enhances the retrieval rate and fitness of the solutions. They also employ heuristic distance measures to computationally optimize the process. These improvements demonstrate the consistency and robustness of the upgraded framework when applied to synthetically generated networks.
The authors then apply their framework to three empirical datasets: subway networks of major cities, regions of street networks, and semantic co-occurrence networks of literature in Artificial Intelligence. The results showcase the capability of the approach to obtain interpretable and decentralized growth processes from these complex networks.
Overall, this research significantly contributes to the field of network modeling by introducing a novel approach that combines genetic programming, symbolic regression, and time-varying network analysis. It provides valuable insights into the generative processes underlying complex networks and opens up new possibilities for understanding and predicting their behavior.