On the discovery of events in EEG data utilizing information fusion

One way to tackle brain computer interfaces is to consider event related potentials in electroencephalography, like the well established P300 phenomenon. In this paper a multiple classifier approach to discover these events in the bioelectrical signal and with them whether or not a subject has recognized a particular pattern, is employed. Dealing with noisy data as well as heavily imbalanced target class distributions are among the difficulties encountered. Our approach utilizes partitions of electrodes to create robust and meaningful individual classifiers, which are then subsequently combined using decision fusion. Furthermore, a classifier selection approach using genetic algorithms is evaluated and used for optimization. The proposed approach utilizing information fusion shows promising results (over 0.8 area under the ROC curve).     

Editorial for ADAC issue 3 of volume 14 (2020)

Advances in Data Analysis and Classification -     

Rank-based classifiers for extremely high-dimensional gene expression data

Predicting phenotypes on the basis of gene expression profiles is a classification task that is becoming increasingly important in the field of precision medicine. Although these expression signals are real-valued, it is questionable if they can be analyzed on an interval scale. As with many biological signals their influence on e.g. protein levels is usually non-linear and thus can be misinterpreted. In this article we study gene expression profiles with up to 54,000 dimensions. We analyze these measurements on an ordinal scale by replacing the real-valued profiles by their ranks. This type of rank transformation can be used for the construction of invariant classifiers that are not affected by noise induced by data transformations which can occur in the measurement setup. Our 10 \(\times \) 10 fold cross-validation experiments on 86 different data sets and 19 different classification models indicate that classifiers largely benefit from this transformation. Especially random forests and support vector machines achieve improved classification results on a significant majority of datasets.     

Search heuristics and the influence of non-perfect randomness: examining Genetic Algorithms and Simulated Annealing

Simulated Annealing and Genetic Algorithms are important methods to solve discrete optimization problems and are often used to find approximate solutions for diverse NP-complete problems. They depend on randomness to change their current configuration and transition to a new state. In Simulated Annealing, the random choice influences the construction of the new state as well as the acceptance of that new state. In Genetic Algorithms, selection, mutation and crossover depend on random choices. We experimentally investigate the robustness of the two generic search heuristics when using pseudorandom numbers of limited quality. To this end, we conducted experiments with linear congruential generators of various period lengths, a Mersenne Twister with artificially reduced period lengths as well as quasi-random numbers as the source of randomness. Both heuristics were used to solve several instances of the Traveling Salesman Problem in order to compare optimization results. Our experiments show that both Simulated Annealing and the Genetic Algorithm produce inferior solutions when using random numbers with small period lengths or quasi-random numbers of inappropriate dimension. The influence on Simulated Annealing, however, is more severe than on Genetic Algorithms. Interestingly, we found that when using diverse quasi-random sequences, the Genetic Algorithm outperforms its own results using quantum random numbers.     

On the fusion of threshold classifiers for categorization and dimensionality reduction

We study ensembles of simple threshold classifiers for the categorization of high-dimensional data of low cardinality and give a compression bound on their prediction risk. Two approaches are utilized to produce such classifiers. One is based on univariate feature selection employing the area under the ROC curve as ranking criterion. The other approach uses a greedy selection strategy. The methods are applied to artificial data, published microarray expression profiles, and highly imbalanced data.     

Multi-objective selection for collecting cluster alternatives

Grouping objects into different categories is a basic means of cognition. In the fields of machine learning and statistics, this subject is addressed by cluster analysis. Yet, it is still controversially discussed how to assess the reliability and quality of clusterings. In particular, it is hard to determine the optimal number of clusters inherent in the underlying data. Running different cluster algorithms and cluster validation methods usually yields different optimal clusterings. In fact, several clusterings with different numbers of clusters are plausible in many situations, as different methods are specialized on diverse structural properties. To account for the possibility of multiple plausible clusterings, we employ a multi-objective approach for collecting cluster alternatives (MOCCA) from a combination of cluster algorithms and validation measures. In an application to artificial data as well as microarray data sets, we demonstrate that exploring a Pareto set of optimal partitions rather than a single solution can identify alternative solutions that are overlooked by conventional clustering strategies. Competitive solutions are hereby ranked following an impartial criterion, while the ultimate judgement is left to the investigator.