Genetic Programming Based on Novelty Search
Abstract
Novelty Search (NS) is a unique approach towards search and optimization,
where an explicit objective function is replaced by a measure
of solution novelty. However, NS has been mostly used in evolutionary
robotics, its usefulness in classic machine learning problems has been
unexplored. This thesis presents a NS-based Genetic Programming
(GP) algorithms for common machine learning problems, with the following
contributions. It is shown that NS can solve real-world classification,
clustering and symbolic regression tasks, validated on realworld
benchmarks and synthetic problems. These results are made
possible by using a domain-specific behavior descriptor, related to the
concept of semantics in GP. Moreover, two new versions of the NS algorithm
are proposed, Probabilistic NS (PNS) and a variant of Minimal
Criteria NS (MCNS). The former models the behavior of each solution
as a random vector and eliminates all the NS parameters while reducing
the computational overhead of the NS algorithm; the latter uses a
standard objective function to constrain and bias the search towards
high performance solutions. The thesis also discusses the effects of NS
on GP search dynamics and code growth. Results show that NS can be
used as a realistic alternative for machine learning, and particularly for
GP-based classification.
Origin | Files produced by the author(s) |
---|
Loading...