- title = "On the Homogenization of Data from Two Laboratories Using Genetic Programming",
- author = "Jose Garcia Moreno-Torres and Xavier Llora and David E. Goldberg and Rohit Bhargava",
- publisher = "Springer",
- year = "2009",
- volume = "6471",
- booktitle = "Learning Classifier Systems",
- series = "Lecture Notes in Computer Science",
- editor = "Jaume Bacardit and Will N. Browne and Jan Drugowitsch and Ester Bernad{\'o}-Mansilla and Martin V. Butz",
- isbn13 = "978-3-642-17507-7",
- pages = "185--197",
- keywords = "genetic algorithms, genetic programming",
-
DOI = "
doi:10.1007/978-3-642-17508-4_12",
- bibdate = "2010-11-30",
- bibsource = "DBLP, http://dblp.uni-trier.de/db/conf/iwcls/iwlcs2009.html#Moreno-TorresLGB09",
- abstract = "In experimental sciences, diversity tends to difficult predictive models' proper generalization across data provided by different laboratories. Thus, training on a data set produced by one lab and testing on data provided by another lab usually results in low classification accuracy. Despite the fact that the same protocols were followed, variability on measurements can introduce unforeseen variations that affect the quality of the model. This paper proposes a Genetic Programming based approach, where a transformation of the data from the second lab is evolved driven by classifier performance. A real-world problem, prostate cancer diagnosis, is presented as an example where the proposed approach was capable of repairing the fracture between the data of two different laboratories.",
-
notes = "booktitle IWLCS. prostate cancer. Ninety three trees
per individual!!! C4.5. Context free grammar but rules
seem to be straight forward does it add anything above
Lisp like tree? 'one point crossover' similar to Koza's
sub tree crossover? Tournament size related to log(pop
size). Population size proportional to number of trees.
\cite{harris:thesis} and \cite{bot:2001:EuroGP}.
Re-represent 93 attributes selected on first lab's data
for use by same C4.5 classifier on second lab's data.
(Includes attributes which are not used by the final
C4.5 classifier). This works.
Solves problem but p195 do not yet give 'any useful information'. Problem to complex? Too much redundant information in each of the 14Gigabytes of information?
",
- affiliation = "Department of Computer Science and Artificial Intelligence, Universidad de Granada, 18071 Granada, Spain",
