Fundamentals of testing AI-based systems · Question 45 of 60
A team has checked an image model with 20 test images, all classified correctly, and now wants to report "100% correct". What speaks against this?
2 answers are correct. Hard
Show answer
Answer
- C A statistically significant sample from the relevant operational data distributions must be tested
- D Because of non-determinism, the test suite must be large enough for statistically significant results
Explanation
A single test or a small number of tests cannot reliably reflect the functional correctness of a probabilistic model. The test suite must be large enough to provide statistically significant results and cover a significant sample from the relevant operational data distributions.