Intelligent, Dynamic and Distributed Solutions for Network Management

This work presents an alternative to the centralized approach of computer network management. In network management platforms, we usually find passive managers and the agents who just have the function of collecting data samples composing the environment. They have no autonomy and the network administrator must reach a decision. This work investigates what is the most suitable Artificial Intelligence (AI) technique applicable in the distributed problem solving of computer network management. In addition, it is usual to find in scientific literature, dynamic problems solved by static feedforward neural networks. Using this static approach, explicitly or implicitly the state of the dynamical system must be supplied, leading to a very big neural network and a corresponding longer training time even if the famous backpropagation is used. The AI techniques will be implemented in autonomous agent format which will be able to automate the network management process.