1: \begin{abstract}

2: In many publications, authors showed that chaotic pseudo random numbers generators (PRNGs) may improve performance of the evolutionary algorithms. In this paper, we use two chaotic maps Gingerbread man and Tinkerbell as the chaotic PRNGs instead of the classical PRNG in the algorithm differential evolution. Numbers generated by this maps are normalized to the interval [0,1] by three different methods -- operation modulo, classical normalization and function atan2. The goal is to show that number normalization may affect the differential evolution convergence speed. In the second part of this paper, we have modified well known PRNG Mersenne Twister to generate numbers with the same probability distribution function like the chaotic maps using different schemes of number normalization, where we have tested if the PRNG with the same probability distribution function will reach the same results. The results mentioned below showed that the selected normalization method may improve differential evolution convergence speed, especially in the case of function atan2 and classical normalization. Modified Mersenne Twister reached very different results than chaotic PRNGs with the different number normalization. That means the distribution of the selected (chaotic) PRNG with the number normalization scheme alone does not suffice to improve differential evolution convergence speed and play the key role in this process.

3: 

4: \keywords{Pseudo random number generator, evolutionary algorithms, differential evolution, particle swarm, chaos, Gingerbread man, Tinkerbell}

5: \end{abstract}