Hi,
For ML with deep learning, it seems to involve lots of empirical practice. However, some clients (e.g., federal government) want to see a sample size statement in assessing feasibility of your proposal – I haven’t been able to come up with something convincing for such requested power analysis. As I am into Week 3 talking about data, I appreciate any guidance from the instructors on how to estimate sample sufficiency for ML, especially deep learning approaches.
Thanks,
Fred