Next: Algorithm GUESS
Up: The Speed Prior: A
Previous: Speed Prior
Speed Prior-Based Inductive Inference
Given
, as we observe an initial segment
of some string,
which is the most likely continuation?
According to Bayes,
 |
(5) |
where
is the measure of
, given
. Having
observed
we will predict those
that maximize
.
Which are those? In what follows, we will confirm the intuition that
for
the only probable continuations of
are those with fast programs. The sheer number of ``slowly'' computable
strings cannot balance the speed advantage of ``more quickly'' computable
strings with equal beginnings.
Proof.
Since no program that requires at least
steps for producing
can compute
in a PHASE with number
, we have
Here we have used the Kraft inequality [13]
to obtain a rough upper bound for the enumerator:
when no
is prefix of another one, then
Q.E.D.
Hence, if we know a rather fast finite program
for
, then Theorem
1 allows for predicting: if we observe some
(
sufficiently large) then it is very unlikely that it was produced by an
-computing algorithm much slower than
.
Among the fastest algorithms for
is FAST itself, which is at
least as fast as
, save for a constant factor. It outputs
after
steps.
Subsections
Next: Algorithm GUESS
Up: The Speed Prior: A
Previous: Speed Prior
Juergen Schmidhuber
2003-02-25
Back to Speed Prior page