Automated Search for the Quantitative Laws Affecting CO2 Fugacity in Sea Water

To describe and explain the world, scientists search for equations that quantify relations among relevant variables. We wanted to assist these efforts by a computer program inspired by genetic programing. Using realworld data, the program discovered equations that outperformed, in terms of accuracy, published expressions. Moreover, the research indicated that somewhat different formulas may be needed in different geographical regions. The latter observation appears to be due to “hidden” parameters that were not included in the available data.

• Type: Conference
• Year: 2007
• Where: FLAIRS
• Authors: Kasun Wickramaratna, Miroslav Kubat, Peter Minnett
• Keywords: Artificial Intelligence, FLAIRS 2007, Genetic Programing, Program Discovered Equations, Relevant Variables