Domain of competence of XCS classifier system in complexity measurement space
read more
Citations
KEEL: a software tool to assess evolutionary algorithms for data mining problems
An overview of ensemble methods for binary classifiers in multi-class problems: Experimental study on one-vs-one and one-vs-all schemes
A study of statistical techniques and performance measures for genetics-based machine learning: accuracy and interpretability
Evolutionary undersampling for classification with imbalanced datasets: Proposals and taxonomy
Evolutionary rule-based systems for imbalanced data sets
References
Genetic algorithms in search, optimization, and machine learning
Reinforcement Learning: An Introduction
Adaptation in natural and artificial systems
Bagging predictors
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the purpose of the search component in XCS?
The search component in XCS is responsible for improving the ruleset, by discovering new promising classifiers and deleting the ones that do not contribute to the knowledge.
Q3. What is the role of the reinforcement component?
The role of the reinforcement component is to evaluate the current classifiers, so that highly fit classifiers correspond to consistent (accurate) descriptions of the target concept.
Q4. What is the reward used to update the parameters of the classifiers in [A]?
Once the action is sent to the environment, the environment returns a reward , which is used to update the parameters of the classifiers in [A].
Q5. how is the xcs classification system based on the generalization hypothesis?
In XCS, this is achieved via a niche GA, by means of: a) the selection operator, which applies local fitness pressure within niches, b) crossover, which is restricted to related classifiers and c) deletion, which tries to balance the size of the niches.
Q6. what causes the evolution of maximally general classifiers?
The fact that the GA takes place in the action sets rather than in the whole population produces a generalization pressure, which leads to the evolution of maximally general classifiers.
Q7. What is the probability of a classifier being deleted?
If one of the classifier’s parents is sufficiently experienced ( & 4 !#"%$ , where7 !#"%$ is a threshold set by the user), accurate ( 5 5$" ) and more general than the classifier, then the classifier is discarded, and the parent’s numerosity is incremented by one.
Q8. What other parameters are used to qualify each classifier?
There are other5 parameters qualifying each classifier, such as: the experience of the classifier (denoted as exp), the average size of the action sets where the classifier has participated (as), the time-step of the last application of the genetic algorithm (ts) and the number of actual micro-classifiers this macroclassifier2 represents, called numerosity (num).
Q9. What is the description of the XCS system?
The codification based on the ternary alphabet (as described in section II-A) has proved to be well suited for a varied range of domains with binary attributes.
Q10. What are the main parameters of a classifier?
Three main parameters estimate the quality of each classifier: a) the payoff prediction 4 , an estimate of the payoff that the classifier will receive if its condition matches the input and its action is selected, b) the prediction error 5 , which estimates the average error between the classifier’s prediction and the received payoff and c) the fitness 6 , an estimate of the accuracy of the payoff prediction.
Q11. Why was subsumption introduced in the first place?
Subsumption was introduced in [2] in order to eliminate some specialized classifiers from the population which were already covered by other accurate and more general classifiers.
Q12. What is the inverse function of the classifier’s error?
the prediction error is updated:5 5 4 5 % (2) Then, the classifier’s accuracy is computed as an inverse function of the classifier’s error:.