Simon M. Garrett

How Do We Evaluate Artificial Immune Systems

The field of Artificial Immune Systems (AIS) concerns the study and development of computationally interesting abstractions of the immune system. This survey tracks the development of AIS since its inception, and then attempts to make an assessment of its usefulness, defined in terms of distinctiveness and effectiveness. In this paper, the standard types of AIS are examinedNegative Selection, Clonal Selection and Immune Networksas well as a new breed of AIS, based on the immunological danger theory. The paper concludes that all types of AIS largely satisfy the criteria outlined for being useful, but only two types of AIS satisfy both criteria with any certainty.

A Link-Based Approach to the Cluster Ensemble Problem

Cluster ensembles have recently emerged as a powerful alternative to standard cluster analysis, aggregating several input data clusterings to generate a single output clustering, with improved robustness and stability. From the early work, these techniques held great promise; however, most of them generate the final solution based on incomplete information of a cluster ensemble. The underlying ensemble-information matrix reflects only cluster-data point relations, while those among clusters are generally overlooked. This paper presents a new link-based approach to improve the conventional matrix. It achieves this using the similarity between clusters that are estimated from a link network model of the ensemble. In particular, three new link-based algorithms are proposed for the underlying similarity assessment. The final clustering result is generated from the refined matrix using two different consensus functions of feature-based and graph-based partitioning. This approach is the first to address and explicitly employ the relationship between input partitions, which has not been emphasized by recent studies of matrix refinement. The effectiveness of the link-based approach is empirically demonstrated over 10 data sets (synthetic and real) and three benchmark evaluation measures. The results suggest the new approach is able to efficiently extract information embedded in the input clusterings, and regularly illustrate higher clustering quality in comparison to several state-of-the-art techniques.

Improved Pattern Recognition with Artificial Clonal Selection

In this paper, we examine the clonal selection algorithm CLONALG and the suggestion that it is suitable for pattern recognition. CLONALG is tested over a series of binary character recognition tasks and its performance compared to a set of basic binary matching algorithms. A number of enhancements are made to the algorithm to improve its performance and the classification tests are repeated. Results show that given enough data CLONALG can successfully classify previously unseen patterns and that adjustments to the existing algorithm can improve performance.

Evolving Controllers for Real Robots: A Survey of the Literature

For many years, researchers in the field of mobile robotics have been investigating the use of genetic and evolutionary computation (GEC) to aid the development of mobile robot controllers. Alongside the fundamental choices of the GEC mechanism and its operators, which apply to both simulated and physical evolutionary robotics, other issues have emerged which are specific to the application of GEC to physical mobile robotics. This article presents a survey of recent methods in GEC-developed mobile robot controllers, focusing on those methods that include a physical robot at some point in the learning loop. It simultaneously relates each of these methods to a framework of two orthogonal issues: the use of a simulated and/or a physical robot, and the use of finite, training phase evolution prior to a task and/or lifelong adaptation by evolution during a task. A list of evaluation criteria are presented and each of the surveyed methods are compared to them. Analyses of the framework and evaluation criteria suggest several possibilities; however, there appear to be particular advantages in combining simulated, training phase evolution (TPE) with lifelong adaptation by evolution (LAE) on a physical robot.

A Link-Based Cluster Ensemble Approach for Categorical Data Clustering

Although attempts have been made to solve the problem of clustering categorical data via cluster ensembles, with the results being competitive to conventional algorithms, it is observed that these techniques unfortunately generate a final data partition based on incomplete information. The underlying ensemble-information matrix presents only cluster-data point relations, with many entries being left unknown. The paper presents an analysis that suggests this problem degrades the quality of the clustering result, and it presents a new link-based approach, which improves the conventional matrix by discovering unknown entries through similarity between clusters in an ensemble. In particular, an efficient link-based algorithm is proposed for the underlying similarity assessment. Afterward, to obtain the final clustering result, a graph partitioning technique is applied to a weighted bipartite graph that is formulated from the refined matrix. Experimental results on multiple real data sets suggest that the proposed link-based method almost always outperforms both conventional clustering algorithms for categorical data and well-known cluster ensemble techniques.

Refining Pairwise Similarity Matrix for Cluster Ensemble Problem with Cluster Relations

Cluster ensemble methods have recently emerged as powerful techniques, aggregating several input data clusterings to generate a single output clustering, with improved robustness and stability. This paper presents two new similarity matrices, which are empirically evaluated and compared against the standard co-association matrix on six datasets (both artificial and real data) using four different combination methods and six clustering validity criteria. In all cases, the results suggest the new link-based similarity matrices are able to extract efficiently the information embedded in the input clusterings, and regularly suggest higher clustering quality in comparison to their competitor.

The balance between initial training and lifelong adaptation in evolving robot controllers

A central aim of robotics research is to design robots that can perform in the real world; a real world that is often highly changeable in nature. An important challenge for researchers is therefore to produce robots that can improve their performance when the environment is stable, and adapt when the environment changes. This paper reports on experiments which show how evolutionary methods can provide lifelong adaptation for robots, and how this evolutionary process was embodied on the robot itself. A unique combination of training and lifelong adaptation are used, and this paper highlights the importance of training to this approach.

Dynamic Function Optimisation: Comparing the Performance of Clonal Selection and Evolution Strategies

This paper reports on novel work using clonal selection (CS) for dynamic function optimisation. A comparison is made between evolution strategies (ES) and CS, for the optimisation of two significantly different dynamic functions at 2, 5 and 10 dimensions. Firstly a sensitivity analysis was performed for both the CS and the ES for both fitness functions. Secondly the performance of the two algorithms was compared over time. The main finding of this work is that the CS optimises better than the ES in problems with few dimensions, although the ES optimises more slowly. At higher dimensions however, the ES optimises both more quickly and to a better level.

A Paratope Is Not an Epitope: Implications for Immune Network Models and Clonal Selection

Artificial Immune Systems (AIS) research into clonal selection and immune network models has tended to use a single, real-valued or binary vector to represent both the paratope and epitope of a B-cell; in this paper, the use of alternative representations is discussed. A theoretical generic immune network (GIN) is presented, that can be used to explore the network dynamics of several families of different B-cell representations at the same time, and that combines features of clonal selection and immune networks in a single model.

Evaluating theories of immunological memory using large-scale simulations

Immunological simulations offer the possibility of performing high-throughput experiments in silico that can predict, or at least suggest, in vivo phenomena. In this paper, we first validate an experimental immunological simulator, developed by the authors, by simulating several theories of immunological memory with known results. We then use the same system to evaluate the predicted effects of a theory of immunological memory. The resulting model has not been explored before in artificial immune systems research, and we compare the simulated in silico output with in vivo measurements. We conclude that the theory appears valid, but that there are a common set of reasons why simulations are a useful support tool, not conclusive in themselves.