Optimization Track¶
This track focuses on improving model and policy quality under resource and feedback constraints.
Recommended Sequence¶
projects/automl-hpo-showcaseprojects/rl-bandits-policy-showcaseprojects/sota-supervised-learning-showcase(benchmark extension)
Core Skills Covered¶
- Hyperparameter search strategy tradeoffs.
- Experiment logging and comparability.
- Bandit policy reward/regret analysis.
- Choosing optimization objectives aligned with business constraints.
Evidence Artifacts To Inspect¶
artifacts/hpo/trials.csvartifacts/hpo/strategy_comparison.csv- reward/regret outputs in RL bandits artifacts
- experiment logs in
artifacts/experiments/
Suggested Reflection Prompts¶
- Which search strategy wins under strict runtime budget?
- How sensitive are recommendations to objective choice?
- What offline checks are required before online policy rollout?