Lightning Talk: Uplink Interference Optimizer, How to Optimize a Cellular Network...
summary
A pretrained graph neural network built with PyTorch and PyTorch Geometric models uplink SINR and lets parameter search run offline so live networks never need weeks of RL. The model uses one-layer GNN with PyTorch Geometric over cell neighbors, predicts SINR with <1% error, and the optimizer tweaks B0 and alpha power-control using genetic search. Trained on a week of aggregated data per site and cross‑site datasets, it outputs single‑shot parameter sets that transfer to subsequent weeks.
AI-written summary · about
chapters (7)
- 0:00Introduction to cell networks
- 1:37Network optimization strategy
- 3:00Graph neural network framework
- 3:43Uplink interference solutions
- 5:07Modeling and data usage
- 7:13Optimization implementation
- 8:28Q&A and future outlook
related talks
14mLightning Talk: Every Millisecond Counts: The Fine-tuning Journey of an Ultra-Eff...
35mClient Side Deep Learning Optimization with PyTorch
18mGeminet: Learning the Duality-based Topology-Agnostic Update Operator for Lightweight...
17mTopoOpt: Co-optimizing Network Topology and Parallelization Strategy for Distributed...
12mLightning Talk: Backpropagation-Free Optimization in PyTorch
40mOptimize Your AI Cloud Infrastructure: A Hardware Perspective
13mOptimizing Dynamic Neural Networks with Brainstorm
01h 01mML for discrete optimization: Theoretical foundations