Algorithmic Information and Randomness
Martin-Löf's 1966 discovery that the theory of computing
and only the theory of computing provides a satisfactory definition of the
randomness of individual binary sequences was an
early and dramatic indicator of things to come.
In addition to providing hardware and software tools, computer science
has developed deep theoretical tools that are conceptually
transforming the foundations of several areas of science.
Many of these tools, like randomness, arise from the growing
unification
of theoretical computer science with information theory.
I have recently been involved in the development of three
types of effective fractal dimension outside of complexity theory:
-
Constructive dimension (2000)
-
Finite-state dimension (2001, with Dai, Lathrop, and Mayordomo)
-
Effective strong dimension (2003, with Athreya, Hitchcock, and Mayordomo)
The first of these refines randomness and gives a new
characterization of Kolmogorov complexity. The second sheds new light
on normal sequences and data compression. The third shows that
packing dimension (an ostensibly deeper fractal dimension
developed in 1982) is an exact dual of Hausdorff dimension with
similarly useful effectivizations.
I am currently working with colleagues and students to extend these developments,
with applications to fractal geometry, dynamical systems, prediction, game theory, and, yes, computational complexity.
Bibliographies
John Hitchcock maintains online bibliographies on resource-bounded
measure and effective
fractal dimensions, with links to available papers in the latter.