Uh... "high memory requirements" means about 2500 bytes. On a laptop with 16gb ram, that's about 1.6e_7 of the available memory.
I kind of suspect that we have more important concerns to be dealing with. (And all psuedo random number generation algorithms suffer from potential predictability issues, among other things.) That said, this -- assertions about a problem which pretty much ignore the scale and importance of the problem -- seems to be quite popular nowadays. Still, ... probably worth thinking about, ...occasionally... -- Raul On Fri, Nov 20, 2020 at 7:40 PM Devon McCormick <[email protected]> wrote: > > This article - https://arxiv.org/pdf/1910.06437.pdf - claims the MT has > some bad properties. One claim I understand is that its extremely long > period length is overkill and this comes with the expense of high memory > requirements. > > The example, which I follow only vaguely, carries the more serious > implication of MT introducing bias into certain kinds of tests. > > Fortunately J gives us a good set of choices of RNGs but I'd be interested > to hear what others think about the charges against MT. > > -- > > Devon McCormick, CFA > > Quantitative Consultant > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
