Pseudo Random Number Generation Using LSTMs and Irrational Numbers
Young-Seob Jeong, Kyo-Joong Oh, Chung-Ki Cho, Ho‐Jin Choi · 2018
Data-driven approaches employ the stochastic process that is simulated using random numbers. The random number can be thought of as an unpredictable value without bias or correlation, so it is essentially impossible to design a system that generates `real' random numbers. There have been studies, therefore, that aimed at developing a pseudo random number generator, where the pseudo random number is not perfectly random, but it is practically useful. In this paper, we propose a new system for pseudo random number generation. The recurrent neural networks with long short-term memery (LSTM) units are used to mimic the appearance of a given sequence of irrational numbers (e.g., pi), and they are supposed to generate pseudo random numbers in iterative manner. We design an algorithm to make the output sequence to have no repetition or pattern, and prove its effectiveness by experimental results.