In this talk, we present a new methodology for the analysis of algorithms.
"Informativeness" is proposed as an alternative measure of algorithmic
performance besides time- and space requirements.
The generalization capability of algorithms w.r.t. to their task of processing
input data to meaningful outputs is introduced. The trade-off between the
informativeness of algorithmic procedures and their stability against
uncertainties is investigated by generalizing Shannon's channel capacity to
arbitrary learning problems.
We exemplify the principle with sorting algorithms. Different algorithms for
the same data processing task, e.g. the various sorting methods, can now be
objectively compared based on the bit rate of their respective capacities.