一种基于随机映射的网络状态评估方法

A network status evaluation method based on random projection

近年来,机器学习技术在网络管理领域得到了广泛使用。然而由于通信网络日益复杂,网络中的非线性和不确定因素使得机器学习变得十分困难。为了提升机器学习的效果,提出了一种采用随机映射的人工神经网络方案,其特点是引入机器学习和网络拓扑的随机性,使得神经网络对学习目标具有更大的适应性,并实现更快、更精确的收敛。相关成果已经在中国移动通信集团山西有限公司(以下简称山西移动)的实际网络中得到了应用并取得较好的效果。

In recent years, machine learning has been widely used in network management. However, the complexity of the communication network is increasing, nonlinear and uncertain factors in the network make the machine learning more difficult. In order to improve the effect of machine learning, a scheme of artificial nervous network based on random projection was proposed. The characteristic of such scheme was the introduction of randomness in machine learning and network topology, which took the learning process more adaptability, and achieve faster and more accurate convergence.