Topic Wishlist

[feamster]

• Tools. Familiarity with PyTorch, AutoGluon, nPrint. (assignments)
• Applications. An overview of the applications of ML4Net and the baselines for typical scenarios, e.g., BBA, Pensieve, Fugu in ABR. (lecture/reading)
• Debugging. More techniques in debugging / interpreting ML4Net or ML models in general. Including permutation-based feature importance, SHAP, LIME, etc. (lecture/reading)
• ML Pipeline. An overview of the full pipeline of operational ML (MLOps) in networking, including data collection, featurization, training, testing, deployment, serving, monitoring, adaptation. (lecture/assignment)
• System Cost. Approaches to measure the system cost of ML4Net models, like some CPU/GPU/Memory measurement tools and approaches to compare the trade-offs. (lecture/assignment)