Narrow Results

This paper presents a structural distance-based crossover for neural network classifiers, which is applied as part of a Memetic Algorithm (MA) for evolving simultaneously the structure and weights of neural network models applied to multiclass problems. Previous researchers have shown that this simultaneous evolution is a way to avoid the noisy fitness evaluation. The MA incorporates a crossover operator that shows to be useful for ameliorating the permutation problem of the network representation (i.e. different genotypes can be used to represent the same neural network phenotype), increasing the structural diversity of the individuals and improving the accuracy of the results. Instead of a recombination probability, the crossover operator considers a similarity parameter (the minimum structural distance), which allows to maintain a trade-off between global and local search. The neural network models selected in this work are the product-unit neural networks (PUNNs), due to their increasing relevance in those classification problems which show a high order relationship between the input variables. The proposed MA is intended to reduce the possible overtraining problems which can raise in some datasets for this kind of models. The evolutionary system is applied to eight classification benchmarks and the results of an analysis of variance contrast (ANOVA) show the effectiveness of the structural-based crossover operator and the capacity of our algorithm to obtain evolved PUNNs with a higher classification accuracy than those obtained using other evolutionary techniques. On the other hand, the results obtained are compared with popular effective machine learning classification methods, resulting in a competitive performance.

In this paper a family of new methods are proposed for constructing hierarchical-interpolative fuzzy rule bases in the frame of a fuzzy rule based supervised machine learning system modeling black box systems defined by input-output pairs. The resulting hierarchical rule base is constructed by using structure building pure evolutionary and memetic techniques, namely, Genetic and Bacterial Programming Algorithms and their memetic variants containing local search steps.

Applying hierarchical-interpolative fuzzy rule bases is a rather efficient way of reducing the complexity of knowledge bases, whereas evolutionary methods (including memetic techniques) ensure a relatively fast convergence in the learning process.

As it is presented in the paper, by applying a newly proposed representation schema these approaches can be combined to form hierarchical-interpolative machine learning systems.

Modern methods for network analytics provide an opportunity to revisit preconceived notions in the classification of diseases as “clusters of symptoms”. Curated collections which were subsequently modified, like the Diagnostic and Statistical Manuals of Mental Disorders “DSM-IV” and the most recent addition, DSM-5 allow us to introspect, using the solution provided by modern algorithms, if there exists a consensus between the clusters obtained via a data-driven approach, with the current classifications. In the case of mental disorders, the availability of a follow-up consensus collection (e.g. in this case the DSM-5), potentially allows investigating if the classification of disorders has moved closer (or away) to what a data-driven analytic approach would have unveiled by objectively inferring it from the data of DSM-IV.

In this contribution, we present a new type of mathematical approach based on a global cohesion score which we introduce for the first time for the identification of communities of symptoms. Different from other approaches, this combinatorial optimization method is based on the identification of “triangles” in the network; these triads are the building block of feedback loops that can exist between groups of symptoms. We used a memetic algorithm to obtain a collection of highly connected-cohesive sets of symptoms and we compare the resulting community structure with the classification of disorders present in the DSM-IV.

In many engineered and natural networked systems, there has been great interest in leader selection and/or edge assignment during the optimal design of controllable networks. In this paper, we present our pioneering work in leader–follower network design via memetic algorithms, which focuses on minimizing the number of leaders or the amount of control energy while ensuring network controllability. We consider three problems in this paper: (1) selecting the minimum number of leaders in a pre-defined network with guaranteed network controllability; (2) selecting the leaders in a pre-defined network with the minimum control energy; and (3) assigning edges (interactions) between nodes to form a controllable leader–follower network with the minimum control energy. The proposed framework can be applied in designing signed, unsigned, directed, or undirected networks. It should be noted that this work is the first to apply memetic algorithms in the design of controllable networks. We chose memetic algorithms because they have been shown to be more efficient and more effective than the standard genetic algorithms in solving some optimization problems. Our simulation results provide an additional demonstration of their efficiency and effectiveness.

As a basic Aristotle element, water is the most abundant and more importantly crucial substance on earth. Without water, there would not be any form of life as we know. Understanding many phenomena in water such as water evaporation and ice melting and formation requires a deep understanding of hydrogen bond breaking and formation. In particular transition states play a key role in the understanding of such hydrogen bond behavior. Transition states, unlike other metastable states, are energy maxima along the minimum energy path connecting two isomers of molecular clusters. Geometry optimization of transition state structures, however, is a difficult task, and becomes even more arduous, especially when dealing with complex biochemical systems using first-principles calculations. In this paper, a novel molecular memetic algorithm (MOL-MA) composing of specially designed molecular-based water evolutionary operators coupled with a transition-state-local search solver and valley adaptive clearing scheme for the discovery of multiple precise transition states structures is proposed. The transition states of water clusters up to four water molecules uncovered using MOL-MA are reported. MOL-MA is shown not only to reproduce previously found transition states in water clusters, but also established newly discovered transition states for sizes 2–4 water molecules. The search performance of MOL-MA is also shown to outperform its compeers when pitted against those reported in the literature for finding transition states as well as recent advances in niching algorithms in terms of solution precision, computational effort, and number of transition states uncovered.

With the advent of social media in our lives and the transformation of consumer behaviour through the impact of the internet technology, online brand–human interactions are crucial in the consumer decision-making process, as well as on corporate performance. In this research study, we have developed a model to predict brand engagement as measured in terms of the amount of earned attention by the brand from the consumer. The exogenous variables of the model comprise parameters related to online traffic, flow of consumer-initiated brand commentaries as measured by three different time frames and the quantity of brand mentions. To test and validate the research model, we have applied a Memetic Algorithm (MA) which is well tailored to the phenomenon of propagation and social contagion. In order to assess this evolutionary algorithm, we have used two alternative local search methods—a Specific Local Search Procedure (SLSP) and the Steepest Ascent (SA) heuristic.We have found that the MA outperforms both local search-based methods, and the SLSP outperforms the SA.