Symbolic Regression Workshop @ GECCO 2012: Program announced

We are glad to announce the program for the Fourth GECCO Workshop on Symbolic Regression and Modeling in Philadelphia. The program for the conference is not announced yet, so the time slots and the exact date will be posted later, but the the talks and abstracts are known:

1. "Graphical Models and What they Reveal about GP When it solves a Symbolic Regression Problem" by Erik Hemberg, Kalyan Veeramachaneni and Una-May O'Reilly

Abstract: We introduce the notion of using graphical models as a new and complementary means of understanding genetic programming dynamics (along with statistics such as mean tree size, etc).  Graphical models reveal the dependency structure of the multivariate distribution associated with functions and terminals in solution structures.  This information is more semantically rather than syntax oriented.  As a first step, using the Pagie-2D problem as our exemplar, we present the generation and inter-generation dynamics of genetic programming in terms of graphical models that are largely unrestricted in structure. Open for discussion are questions such as: should a estimation of distribution  genetic programming algorithm mimic standard  genetic programming's search bias in terms of tree size and shape? And, does graphical model analysis indicate a better way to control the search bias for symbolic regression - by operator design, size control, bloat control or other means?

2. "Sequential Parameter Optimization for Symbolic Regression" by Thomas Bartz-Beielstein, Oliver Flasch, and Martin Zaefferer

Abstract: Modern Symbolic Regression (SR) engines are complex systems of many components, most of which require some form of parameterization. In this talk, we show how to apply Sequential Parameter Optimization (SPO) as a rigorous method for finding near-optimal parameter settings for SR systems. As modern SR systems often offer alternative operator sets for population initialization, variation, and selection, we also demonstrate how to use modern Design of Experiments (DoE) methods to find problem-specific near-optimal SR system configurations, in addition to near-optimal parameterizations for each selected system component. The experimental design for SR can somehow be tricky, because of interactions in the parameter settings. Methods for handling configurations of parameters which depend on higher-level parameters will be presented. Our exposition is based on a simple framework for statistical sound, reproducible empirical research in SR.

3. "Robust function discovery and feature selection with symbolic regression for life sciences and engineering" by Mark Kotanchek and Katya Vladislavleva

Abstract: Industrial process and product optimization is impossible without meaningful models and insights on significant features controlling process or product performance. Real-world modeling and feature selection problems have many issues - high-dimensional, non-linear, with unbalanced measurements, correlated features, missing experiments, etc., which makes it difficult for most people to know what the right approach is in any given situation. We present a function discovery technology based on symbolic regression that routinely converts these problems  into meaningful and insightful models with robust driver features identification. Without requiring a Ph.D. in Computer Science or Statistics, it is now possible to easily develop robust nonlinear models (complete with trust measures), identify data outliers and interactively explore the model dynamics and response sensitivities. Our presentation will illustrate the ease and power of automatic conversion of a spreadsheet of data into an interactive data story report using examples drawn from life sciences and engineering.

4."A New Framework for Scalable Genetic Programming" by Nassima Aleb and Samir Kechid

We present a new framework for scalable and efficient multi-objective genetic programming. We introduce a new program modeling aiming at facilitating programs creation, execution and improvement. The proposed modeling dramatically reduces the time of programs executions and allows well founded programs recombination.