#fitness-function

Question 1: Single choice

A genetic algorithm evolves randomly initialized neural networks that should recognize a secret knock pattern, given as a 0/1 sequence $k_1 \ldots k_n$ . Which fitness function would leave the algorithm stuck on a fitness plateau from the very first generation?

The number of positions $k_i$ that are classified correctly1
The number of correctly classified subsequences $k_i \ldots k_j$ 2
1 if the whole pattern is classified correctly, 0 otherwise3
The length of the prefix $k_1 \ldots k_t$ that is classified correctly4

Timed test