Session details: Learning Classifier systems and other genetics-based machine learning: posters

A byproduct benefit of using probabilistic model-building genetic algorithms is the creation of cheap and accurate surrogate models. Learning classifier systems -- and genetics-based machine learning in general -- can greatly benefit from such surrogates which may replace the costly matching procedure of a rule against large data sets. In this paper we investigate the accuracy of such surrogate fitness functions when coupled with the probabilistic models evolved by the $\chi$-ary extended compact classifier system ($\chi$eCCS). To achieve such a goal, we show the need that the probabilistic models should be able to represent all the accurate basis functions required for creating an accurate surrogate. We also introduce a procedure to transform populations of rules based into dependency structure matrices (DSMs) which allows building accurate models of overlapping building blocks -- a necessary condition to accurately estimate the fitness of the evolved rules.

Learning Classifier System (LCS) finds a new rule set using genetic algorithm (GA). In this paper, The Zeroth Level Classifier System (ZCS) is applied to evolving the strategy of a robot soccer simulation game (SimuroSot), which is a state varying dynamical system changed over time, as GBML (Genetic Based Machine Learning) and we show the effectiveness of the proposed scheme through the simulation of robot soccer.

Rough set theory has proved efficient for many applications, including finding hidden patterns in data, data reduction, evaluating significance of data, and generating sets of decision rules from data. Recently, it has shown to be effective approach for constructing ensemble learning systems as well. Learning classifier systems are genetics-based machine learning techniques that have recently shown a high degree of competence on a variety of data mining problems. Attempts to improve its generalization capabilities in the literature using ensemble learning lack a systematic and robust techniques for partitioning the problem at hand. It is well known that ensemble performance depends on the problem decomposition technique being used. Rough set based learning classifier system ensemble is proposed in this paper. In this approach, rough set attribute reduction is used to generate a set of reducts, and then a diverse subset of these reducts is selected to train an ensemble of base classifiers. The experiments show that classification accuracy of reduct-based ensemble systems outperforms a single learning classifier system model. It has also shown better performance than either an ensemble of classifiers with all attributes being used or a single classifier trained by a single reduct. It has also shown competitive performance to the random subspace ensemble strategy on the set of real data sets used in the experiments.

A learning classifier system is one of the methods for applying a genetic-based approach to machine learning applications. An enhanced version of the system that employs the Bucket-brigade algorithm to reward individuals in a chain of co-operating rules is implemented and assigned the task of learning rules for control robotics behaviors. Illustrates the approach with the example of kicking a moving Ball into a goal (KMB). The KMB (kick a Moving Ball) System built of two-classifier subsystems work together, each classifier system learns a simple behavior, first classifier system, learn simulated robot chase behavior i.e. learn robot to move single step toward moving ball, second classifier system, learn the simulated robot approach behavior i.e. learn robot to kick the ball toward fixed goal, the system as a whole has as its learning goal the coordinate of behaviors. This work examine the performance of the simple classifier system (SCS) on the (KMB) problem, perform two SCS simulations one without the genetic algorithm enabled (GA) and one with the genetic algorithm enabled (GA) results using a classifier system with genetic algorithm show improvement over one without, and furthermore the level of performance has been high enough to rival human accuracy. Also the results without the genetic algorithm show that the apportionment of credit algorithm adjust the strength values of the rules.

Data mining, and specifically supervised data classification, is a key application area for Learning Classifier Systems (LCS). Scaling to larger classification problems, especially to higher dimensional problems, is a key challenge. Ensemble based approaches can be applied to LCS to address scalability issues. To this end a rough set based ensemble of LCS is proposed, which relies on a pre-processed feature partitioning step to train multiple LCS on feature subspaces. Each base classifier in the ensemble is a Michigan style supervised LCS. The traditional genetic algorithm based rule evolution is replaced by a differential evolution based rule discovery, to improve generalisation capabilities of LCS. A voting mechanism is then used to generate output for test instances. This paper describes the proposed ensemble algorithm in detail, and compares its performance with different versions of base LCS on a number of benchmark classification tasks. Analysis of computational time and model accuracy show the relative merits of the ensemble algorithm and base classifiers on the tested data sets. The rough set based ensemble learning approach and differential evolution based rule searching out-perform the base LCS on classification accuracy over the data sets considered. Results also show that small ensemble size is sufficient to obtain good performance.

As a broad subfield of artificial intelligence, machine learning is concerned with the development of algorithms and techniques that allow computers to learn. These methods such as fuzzy logic, neural networks, support vector machines, decision trees and Bayesian learning have been applied to learn meaningful rules; however, the only drawback of these methods is that it often gets trapped into a local optimal. In contrast with machine learning methods, a genetic algorithm (GA) is guaranteeing for acquiring better results based on its natural evolution and global searching. GA has given rise to two new fields of research where global optimization is of crucial importance: genetic based machine learning (GBML) and genetic programming (GP). This article adopts the GBML technique to provide a three-phase knowledge extraction methodology, which makes continues and instant learning while integrates multiple rule sets into a centralized knowledge base. Moreover, the proposed system and GP are both applied to the theoretical and empirical experiments. Results for both approaches are presented and compared. This paper makes two important contributions: (1) it uses three criteria (accuracy, coverage, and fitness) to apply the knowledge extraction process which is very effective in selecting an optimal set of rules from a large population; (2) the experiments prove that the rule sets derived by the proposed approach are more accurate than GP.

We present a bibliography of all works we could find on Learning Classifier Systems (LCS) - the genetics-based machine learning systems introduced by John Holland. With over 400 entries, this is at present the largest bibliography on classifier systems in existence. We include a list of LCS resources on the world wide web.

We propose a parallel distributed model of a hybrid fuzzy genetics-based machine learning (GBML) algorithm to drastically decrease its computation time. Our hybrid algorithm has a Pittsburgh-style GBML framework where a rule set is coded as an individual. A Michigan-style rule-generation mechanism is used as a kind of local search. Our parallel distributed model is an island model where a population of individuals is divided into multiple islands. Training data are also divided into multiple subsets. The main feature of our model is that a different training data subset is assigned to each island. The assigned training data subsets are periodically rotated over the islands. The best rule set in each island also migrates periodically. We demonstrate through computational experiments that our model decreases the computation time of the hybrid fuzzy GBML algorithm by an order or two of magnitude using seven parallel processors without severely degrading the generalization ability of obtained fuzzy rule-based classifiers. We also examine the effects of the training data rotation and the rule set migration on the search ability of our model.

Various evolutionary multiobjective optimization (EMO) algorithms have been used in the field of evolutionary fuzzy systems (EFS), because EMO algorithms can easily handle multiple objective functions such as the accuracy maximization and complexity minimization for fuzzy system design. Most EMO algorithms used in EFS are Pareto dominance-based algorithms such as NSGA-II, SPEA2, and PAES. There are a few studies where other types of EMO algorithms are used in EFS. In this paper, we apply a multiobjective evolutionary algorithm based on decomposition called MOEA/D to EFS for fuzzy classifier design. MOEA/D is one of the most well-known decomposition-based EMO algorithms. The key idea is to divide a multiobjective optimization problem into a number of single-objective problems using a set of uniformly distributed weight vectors in a scalarizing function. We propose a new scalarizing function called an accuracy-oriented function (AOF) which is specialized for classifier design. We examine the effects of using AOF in MOEA/D on the search ability of our multiobjective fuzzy genetics-based machine learning (GBML). We also examine the synergy effect of MOEA/D with AOF and parallel distributed implementation of fuzzy GBML on the generalization ability.

This issue is dedicated to Stewart Wilson both in recognition of his achievements and in appreciation for his guidance of the field. Not only has his research transformed the field, but his support, encouragement, and advice have been invaluable to many of us individually and to the community as a whole. Learning Classifier Systems (LCS) are rule-based machine learning algorithms introduced by Holland (see [2, 4–6]) that learn a population of IF-THEN rules that specify ‘‘IF x happens THEN do (or predict) y’’. In 1995 Stewart Wilson, already a leading LCS researcher, published a paper entitled ‘‘Classifier Fitness based on Accuracy’’ [16] that introduced the XCS algorithm. This paper proved a turning point for the field as XCS and its derivatives rapidly became the main focus of LCS research, and continue to be so today. We organised this special issue of Evolutionary Intelligence to mark the 20th anniversary of this landmark paper, and to serve as post-proceedings for IWLCS 2014: the Seventeenth International Workshop on Learning Classifier Systems. The annual IWLCS meetings are the yearly highlight of the LCS calendar and have no doubt contributed to the strong sense of community in the LCS field. For 2015 IWLCS has been rebranded as the more descriptive IWERML: the International Workshop on Evolutionary Rule-based Machine Learning. In 1994, the year before introducing XCS, Wilson introduced ZCS, the ‘‘Zeroth-order Classifier System’’ [15]. Wilson felt the growing complexity of the LCS concept was hindering progress and ZCS was an attempt to strip the LCS down to its essentials while retaining a working system. XCS built on the minimalist ZCS with two radical changes: a switch to accuracy-based fitness and, based on work by Booker [1], the addition of a niche Genetic Algorithm (GA) that strongly favours general rules. This combination was a hit: the niche GA favours general rules but accuracy-based fitness insists strictly that they make good predictions. This combination drives XCS to learn rules that are as general as possible while remaining accurate. Bull’s article in this issue provides much more on the history of LCS prior to and following XCS [2]. The famous ‘‘accuracy-based fitness’’ of XCS needs some explanation. The earlier ZCS featured ‘‘strengthbased fitness’’, in which the fitness of a rule in the GA was derived from its strength, a measure of the amount of reward the rule received. Consequently, fit rules in ZCS are those that receive a lot of reward. In contrast, XCS rules are fit if they make consistently accurate predictions about the reward they receive. This has the counterintuitive consequence that XCS rules that consistently take a bad action can be fit, since they are consistent. However, this poses no problem when it comes time for XCS to choose an action, since that can be done using the magnitude of the reward predicted by each rule. Why does XCS retain rules whose action it does not use? This ‘‘complete map’’ of the state/ & Muhammad Iqbal muhammad.iqbal@ecs.vuw.ac.nz

We propose that the behavior of nonlinear media can be controlled automatically through evolutionary learning. By extension, forms of unconventional computing (viz., massively parallel nonlinear computers) can be realized by such an approach. In this initial study a light-sensitive subexcitable Belousov-Zhabotinsky reaction in which a checkerboard image, composed of cells of varying light intensity projected onto the surface of a thin silica gel impregnated with a catalyst and indicator, is controlled using a learning classifier system. Pulses of wave fragments are injected into the checkerboard grid, resulting in rich spatiotemporal behavior, and a learning classifier system is shown to be able to direct the fragments to an arbitrary position through dynamic control of the light intensity within each cell in both simulated and real chemical systems. Similarly, a learning classifier system is shown to be able to control the electrical stimulation of cultured neuronal networks so that they display elementary learning. Results indicate that the learned stimulation protocols identify seemingly fundamental properties of in vitro neuronal networks. Use of another learning scheme presented in the literature confirms that such fundamental behavioral characteristics of a given network must be considered in training experiments.

Learning classifier systems (LCSs) have existed for nearly twenty years (Holland & Reitman, 1978). Research efforts in reinforcement learning (RL), evolutionary computation (EC), and neural networks have enhanced the original LCS paradigm. New thoughts from these areas have created a renaissance period for the LCS. This paper highlights some key LCS advancements and the fields that inspired them. One inspiration, from neural networks, is examined for a novel LCS approach to autonomous mobile robots. A simple, LCS-controlled robot simulation is presented. This simulation shows the potential benefits of combined biological paradigms and the hybridization of ideas in the LCS. Future directions for LCS research are discussed.

Learning classifier systems: New models, successful applications

ENKOS – A Smart Home control system basing on learning classifier systems

Learning classifier systems (LCSs) have been successfully adapted to real-world domains with the claim of human-readable rule populations. However, due to the inherent rich characteristic of the employed representation, it is possible to represent the underlying patterns in multiple (polymorphic) ways, which obscures the most informative patterns. A novel rule reduction algorithm is proposed based on ensembles of multiple trained LCSs populations in a hierarchical learning architecture to reduce the local diversity and global polymorphism. The primary aim of this project is to interrogate the hidden patterns in LCSs’ trained population rather than improve the predictive power on test sets. This enables successful visualization of the importance of features in data groups (niches) that can contain heterogeneous patterns, i.e. even if different patterns result in the same class the importance of features can be found.