Pseudorandom functions whose asymptotic distributions are asymptotically Gaussian (Q808089)

Given an ergodic transformation T on the Lebesgue space ([0,1],dx) and a continuous function h(x) \((\int^{1}_{0}h(x)dx=0)\), we are interested in asymptotic behaviours of the pseudorandom functions, \(q_{\theta}(t;x)=\sqrt{\theta}\cdot h(T^{[\theta t]}x)\) (x\(\in (0,1)\), \(\theta >0)\). It is shown that, if T possesses a nice mixing property, the functions behave, as \(\theta\) tends to infinity, like sample paths of the white Gaussian noise, in the following sense: We consider the new pseudorandom functions \(Q_{\theta}(t;x)\) defined by the formula \(Q_{\theta}(t;x)=\int^{\infty}_{-\infty}K(t- s)q_{\theta}(s;x)ds,\) where K(t) is a smooth function with bounded support in \(R^ 1\). Then, for almost all \(x_ 0\) in (0,1), the asymptotic distributions of \(Q_{\theta}(t;x_ 0)\) converge to the same normal distribution \(N(0,\sigma^ 2)\) with \(\sigma^ 2=\int^{\infty}_{-\infty}K^ 2(t)dt\), as \(\theta\to \infty.\) This result is a variant of that due to P. P. Hien [cf., \textit{J. Bass}, ibid. 47, 354-399 (1974; Zbl 0286.76024) and ibid., 458-503 (1974; Zbl 0284.76040)] which was obtained for the case of completely uniformly distributed pseudorandom sequences in place of the family of uniformly distributed pseudorandom sequences \(\{T^ nx\}\). But from the viewpoint of applications, ours are much more ameliorated than the P. P. Hien's result. Since in our case the conditions for the transformation T are satisfied by a wide class of piecewise \(C^ 2\) transformations which generate many important and practical pseudorandom sequences with much weaker ``randomness'' than the complete uniformity, they can even fail to be 2-uniformly distributed.