How are test cases (for testing algorithms), akin to test data for machine learning models?
All models are non-deterministic algorithms. (quote relevant)
Your accuracy must represent the problem at hand, and otherwise you can’t really reliably gauge how ‘accurate’ model is. The data defines your interpretation of the quality of your algorithm/model.
Both algorithms and machine learning algorithms should be made more robust by testing against or feeding more edge cases.
You’ve been hired by Goose Goose Go to implement an ‘auto-complete’ feature for their new search engine. How would you go about this?
- Algorithmic: Just use tries
- Machine learning: