Hold-out Testing
A core method in machine learning and statistics for assessing the reliability of an evaluation. A portion of the data is held out and used only for final validation, taking no part in model training or parameter search.
Timeline (last 90 days)
The author built a simulated test framework (held-out test), implanting known text on a fake plaque and then cracking it with the same method to measure the reliability of the method.
Held-out testing requires that the test data be completely isolated from the search process, otherwise information leakage occurs and leads to overestimating the method's capability.
All Facts (2)
The author built a simulated test framework (held-out test), implanting known text on a fake plaque and then cracking it with the same method to measure the reliability of the method.
50%UnverifiedHeld-out testing requires that the test data be completely isolated from the search process, otherwise information leakage occurs and leads to overestimating the method's capability.
50%