• ML-based Traffic Engineering for Wide-Area-Networks
  • Supervised Learning Mechanisms and Reinforcement Learning for Traffic Engineering
  • Distributed ML-based Traffic Engineering approaches
  • Efficient network representation with Graph Neural Networks
• Network Optimization for Distributed ML Training
  • Application of Traffic Engineering mechanisms to accelerate distributed ML training
  • Improving utilization and training speed of distributed GPU clusters
  • Distributed ML training acceleration techniques: Gradient Compression, Worker Node Synchronization, Collective Communications

KiKIT - The Helmholtz Pilot Program Core-Informatics at KIT