# A Survey on the Network Models applied in the Industrial Network Optimization

**Authors:** Chao Dong, Xiaoxiong Xiong, Qiulin Xue, Zhengzhen Zhang, Kai Niu, Ping Zhang  
**arXiv:** [2209.08294](https://arxiv.org/abs/2209.08294)  
**Date:** 2022-09-17  
**Summary page:** https://www.summarizepaper.com/en/paper/2209.08294v1/

## Key points

- Optimization of industrial networks relies on network architecture design
- Different types of network architectures based on scale: small-scale and large-scale networks
- Graph theory used to model these topologies
- Small-scale networks exhibit regular topology, while large-scale networks have random node and edge distributions
- Popular models for analyzing large-scale networks include random networks, small-world networks, and scale-free networks
- Introduction of a novel metric called "system entropy" to evaluate network performance
- System entropy has properties of non-negativity, monotonicity, and concave-convexity
- System entropy can be applied to regular networks, random networks, small-world networks, and scale-free networks
- Simulation results show that system entropy facilitates comparisons among various industrial networks under different models
- Survey provides valuable insights into network architecture design for optimization in industrial settings.

## Summary

The optimization of industrial networks relies heavily on the design of network architecture. This study explores different types of network architectures based on their scale, categorizing them into small-scale and large-scale networks. To model these topologies, graph theory is employed as an efficient mathematical tool. Small-scale networks typically exhibit regular topology while large-scale networks are characterized by random node and edge distributions. In recent years, various models have gained popularity in the analysis of large-scale networks including random networks, small-world networks, and scale-free networks. This survey delves into these models and provides a comprehensive summary and analysis of their application in industrial scenarios. Additionally, this study introduces a novel metric called "system entropy" to evaluate network performance. The mathematical properties of system entropy are examined from the perspectives of non-negativity, monotonicity, and concave-convexity. One notable advantage of this metric is its ability to encompass regular networks, random networks, small-world networks, and scale-free networks while maintaining strong generality. Simulation results demonstrate that system entropy effectively facilitates comparisons among various industrial networks under different models. With its comprehensive approach to network modeling methods based on graph theory and practical applications in industrial settings, this survey contributes valuable insights to the field of network architecture design for optimization purposes.