Bild mit Unilogo
home uni uni kontakt contact
unilogo Universität Stuttgart
Institute of Communication Networks and Computer Engineering (IKR)

Project description

Druckansicht
 

Master thesis No. 1063    (Offer)   [pdf]

Designing and Implementing Graph Neural Networks for Efficient IP-Optical Networking


Methods

Topics

Graph Neural Networks
Machine Learning
Network Modeling
Simulation

Multi-layer networks
Network control


Description

Background

Today, communication networks have evolved into complex systems handling vast amounts of data, making efficient management and optimization crucial. The network topology and state can be naturally represented as a graph structure, with nodes representing network elements and edges representing connections. Graph Neural Networks (GNNs) have emerged as a powerful tool for learning from graph-structured data, making them particularly suitable for network optimization problems. The Routing and Spectrum Assignment (RSA) problem in IP-optical networks presents a significant challenge that could benefit from GNN-based approaches, as it requires understanding both topology and state information.

Problem Description

In the context of this thesis, you are called to apply GNNs to solve network optimization problems. More specifically, the thesis consists of the following steps:

Familiarize yourself with Julia and related GNN packages

Design appropriate graph representations for network states

Implement and train GNN models for the RSA problem

Compare different GNN architectures and aggregation functions

Evaluate the impact of GNNs on solving the RSA problem

Acquired Knowledge and Skills

In this thesis, you will gain deep understanding of Graph Neural Networks and their application to networking problems. You will experiment with the scientific programming language Julia and its graph learning libraries. Additionally, you will get great insight into networking and network services.


Requirements

Desirable knowledge

Programming Experience
Communication Networks Architecture and Design
Basic Machine Learning Knowledge

Programming Experience in Julia
Graph Theory Basics
Neural Networks Fundamentals


Contact

M.Sc. Nicolas Hornek, room 1.402 (ETI II), phone 685-67992, [E-Mail]

Dipl.-Ing. Filippos Christou, room 1.319 (ETI II), phone 685-67968, [E-Mail]