Abstract: Moraglio et al. have recently introduced new genetic
operators for genetic
programming, called geometric semantic operators. These operators induce
a unimodal fitness landscape for all the problems consisting in matching
input
data with known target outputs (like regression and classification). This
feature
facilitates genetic programming evolvability, which makes these
operators extremely
promising. Nevertheless, Moraglio et al. leave one big open problem:
these operators, by construction, always produce offspring that are
larger than their parents,
causing an exponential growth in the size of the individuals, which
actually renders
them useless in practice.
In this seminar, I offer a general introduction to optimization, to
fitness landscapes
and to evolutionary computation. After that, I present geometric semantic
operators and I show that they induce a unimodal fitness landscape
on every possible instance of regression and classification.
Finally, after discussing the limitation of geometric semantic operators,
I show a new efficient implementation of them, recently proposed
by myself in collaboration with Sara Silva, Mauro Castelli and Luca
Manzoni.
This allows us, for the first time, to use them on complex real-life
applications,
like the two problems in pharmacokinetics that I discuss in the seminar.
The presented experiments confirm the excellent evolvability
of geometric semantic operators, demonstrated by the good results
obtained on
training data. Furthermore, I show that we have also achieved a
surprisingly
good generalization ability, and I discuss the fact it that can be
explained
considering some properties of geometric semantic operators, which makes
them even more appealing than before.