Natural Coding: A More Efficient Representation for Evolutionary Learning

Giráldez, Raúl; Aguilar–Ruiz, Jesús S.; Riquelme, José C.

doi:10.1007/3-540-45105-6_109

Cited by 14 publications

(14 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first one is called Standard-Hider and is the version presented in [9]. The second one is the same algorithm but it incorporates the self-setting of mutation probability according to the previous section.…”

Section: Resultsmentioning

confidence: 99%

“…As we will see in Section 7, two versions of this algorithm will be used: the original algorithm, without changes, and an adapted version to include the feature influence across the evolutionary process. The reader can find a detailed description of Hider in [1,9].…”

Section: Hidermentioning

confidence: 99%

“…Hider encodes each rule (phenotype) as one individual (genotype) by using an original representation, called natural coding [9]. This encoding assigns only one gene for each attribute, regardless of its type (continuous or discrete), so that each gene represents only one condition in the rule.…”

Section: Application: Mutationmentioning

confidence: 99%

See 2 more Smart Citations

Feature influence for evolutionary learning

Giráldez

Aguilar–Ruiz

2005

Proceedings of the 7th Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

This paper presents an approach that deals with the feature selection problem, and includes two main aspects: first, the selection is done during the evolutionary learning process, i.e., it is a dynamic approach; and second, the selection is local, i.e., the algorithm selects the best features from the best space region to learn at a given time of the exploration process. While the traditional feature selection is based on the attribute relevance, our approach is based on a new concept, called feature influence, which is aware of the dynamics and locality of the concept. The feature influence provides a measure of the attribute relevance at a certain instant of the evolutionary learning process, since it depends on each generation. Experimental results have been obtained by comparing an EA-based supervised learning algorithm to its modified version to include the concept approached. The results show an excellent performance, as the new adapted algorithm achieves the same classification results while using less rules, less conditions in rules and much less generations. The experiments include the statistical significance of the improvement over a set of sixteen datasets from the UCI repository.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Hidermentioning

confidence: 99%

See 1 more Smart Citation

Feature influence for evolutionary learning

Giráldez

Aguilar–Ruiz

2005

Proceedings of the 7th Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…Some methods use some kind of discretization [1,21,24] to convert the continuous attributes to categorical ones, which allows the usage of categorical knowledge representations [18,39]. A different alternative is to use rules based on fuzzy logic.…”

Section: Related Workmentioning

confidence: 99%

Improving the scalability of rule-based evolutionary learning

2008

View full text Add to dashboard Cite

Evolutionary learning techniques are comparable in accuracy with other learning methods such as Bayesian Learning, SVM, etc. These techniques often produce more interpretable knowledge than, e.g. SVM; however, efficiency is a significant drawback. This paper presents a new representation motivated by our observations that Bioinformatics and Systems Biology often give rise to very large-scale datasets that are noisy, ambiguous and usually described by a large number of attributes. The crucial observation is that, in the most successful rules obtained for such datasets, only a few key attributes (from the large number of available ones) are expressed in a rule, hence automatically discovering these few key attributes and only keeping track of them contributes to a substantial speed up by avoiding useless match operations with irrelevant attributes. Thus, in effective terms this procedure is performing a fine-grained feature selection at a rule-wise level, as the key attributes may be different for each learned rule. The representation we propose has been tested within the BioHEL machine learning system, and the experiments performed show that not only the representation has competent learning performance, but that it also manages to reduce considerably the system run-time. That is, the proposed representation is up to 2-3 times faster than state-of-the-art evolutionary learning representations designed specifically for efficiency purposes

show abstract

“…In Table XI, we compare the results obtained by ECL on the propositional datasets against the results onbtained by four nonevolutionary systems for ICL, C4.5 [24], Naive Bayes [25], SMO [26], and IB1 [27], and two evolutionary algorithms, HIDER* [28] and GAssist [29]. C4.5 is a decision tree algorithm, Naive Bayes uses the Baye's rule of conditional probabilities to estimate the predicted class, SMO implements the sequential minimal optimization algorithm for training a support vector classifier, and IB1 uses a simple distance measure to find the training instance closest to the given test instance, and predicts the same class as this training instance.…”

Section: Comparison With Other Systemsmentioning

confidence: 99%

Handling Continuous Attributes in an Evolutionary Inductive Learner

Divina

Marchiori

2005

IEEE Trans. Evol. Computat.

View full text Add to dashboard Cite

This paper analyzes experimentally discretization algorithms for handling continuous attributes in evolutionary learning. We consider a learning system that induces a set of rules in a fragment of first-order logic (evolutionary inductive logic programming), and introduce a method where a given discretization algorithm is used to generate initial inequalities, which describe subranges of attributes' values. Mutation operators exploiting information on the class label of the examples (supervised discretization) are used during the learning process for refining inequalities. The evolutionary learning system is used as a platform for testing experimentally four algorithms: two variants of the proposed method, a popular supervised discretization algorithm applied prior to induction, and a discretization method which does not use information on the class labels of the examples (unsupervised discretization). Results of experiments conducted on artificial and real life datasets suggest that the proposed method provides an effective and robust technique for handling continuous attributes by means of inequalities.

show abstract

Natural Coding: A More Efficient Representation for Evolutionary Learning

Cited by 14 publications

References 13 publications

Feature influence for evolutionary learning

Feature influence for evolutionary learning

Improving the scalability of rule-based evolutionary learning

Handling Continuous Attributes in an Evolutionary Inductive Learner

Contact Info

Product

Resources

About