Motivation
For centuries researchers have been interested in automating the
knowledge discovery process.
When the objective is to get insights regarding data sets,
models can be considered among the desirable artifacts produced
by this task.
Traditionally, the form of the model is explicitly defined by a
domain specialist, leaving just coefficients or parameters to be
further adjusted.
However, it is possible to go beyond; in symbolic regression the
structure of the model is no longer predefined by the user but
rather included as part of the problem.
Symbolic regression aims at finding symbolic descriptions,
usually as mathematical expressions, decision rules, or even
programs in a certain language, in order to describe and
communicate new knowledge as well as assist the decision making
process in various domains.
The bio-inspired genetic programming paradigm, originally
designed to find computer programs in arbitrary human-readable
languages, is well-suited and widely applied to symbolic
regression problems.
However, other nature-inspired search engines can potentially be
adopted in symbolic regression.
Scope
The session seeks to promote the presentation and discussion of
innovative techniques for symbolic regression problems, model
inference, and knowledge discovery involving (but not limited
to):
- Evolutionary algorithms
- Swarm intelligence
- Immune Systems
- Physically-inspired techniques
- Novel nature-inspired techniques
- New representations and/or operators
- Any genetic programming (GP) variant
- Parallel and distributed algorithms
- Empirical studies of GP performance and behavior
as well as new applications, specially in real-world problems.