As an illustration of the unfolding of scale invariance in stochastic and s
tatistical mechanical systems we consider some properties of random-walks w
ith long-tailed step distributions. We analyze how a renormalization group
(RG) transformation flows along maximum entropy step distributions under a
given constrained moment and ends at a fixed-point distribution that is ide
ntified as that for the Weierstrass walk. This walk displays a transition f
rom Gaussian to fractal behavior and it is appropriate to employ the genera
lized Tsallis entropy for the latter regime. Along the RG flow the entropy
decreases monotonically and reaches a minimum at the fixed point. (C) 2001
Elsevier Science Ltd. All rights reserved.