Computation of critical distances within multiplicative congruential pseudorandom number sequences (Q1185973)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Computation of critical distances within multiplicative congruential pseudorandom number sequences
scientific article

    Statements

    Computation of critical distances within multiplicative congruential pseudorandom number sequences (English)
    0 references
    0 references
    0 references
    28 June 1992
    0 references
    Let \(y_ 1,y_ 2,y_ 3,\dots\) be a sequence of pseudorandom numbers generated by a multiplicative congruential generator \(G\). The last two authors discovered that there are \(s\) so-called critical shifts such that the points \((y_ n,y_{n+s})\) concentrate on few parallel lines. So \((y_ n,y_{n+s})\), \(n=1,2,3,\ldots\) are strongly correlated and the sequences \(y_ 1,y_ 2,\dots,y_ m\) with \(m>s\) should not be used. The authors present the algorithm computing critical shifts. No detailed proof is presented. The algorithm was used on a generator from IBM library \(y_{n+1}=7^ 5y_ n\) \(\bmod (2^{31}-1)\). The smallest critical shift of it is 642551. The generator \(y_ n=71971110957370*y_{n-1}\) \(\bmod(2^{47}- 115)\) has the least shift greater than \(10^{10}\).
    0 references
    0 references
    critical distances
    0 references
    multiplicative congruential pseudorandom number sequences
    0 references
    random number generators
    0 references
    critical shifts
    0 references