Abstract. Research in aggregation theory is nowadays still mostly focused on algorithms to summarize tuples consisting of observations in some real interval or of diverse general ordered structures. Of course, in practice of information processing many other data types between these two extreme cases are worth inspecting. This contribution deals with the aggregation of lists of data points in Rd for arbitrary d≥1. Even though particular functions aiming to summarize multidimensional data have been discussed by researchers in data analysis, computational statistics and geometry, there is clearly a need to provide a comprehensive and unified model in which their properties like equivariances to geometric transformations, internality, and monotonicity may be studied at an appropriate level of generality. The proposed penalty-based approach serves as a common framework for all idempotent information aggregation methods, including componentwise functions, pairwise distance minimizers, and data depth-based medians. It also allows for deriving many new practically useful tools.
Abstract. Economic inequality measures are employed as a key component in various socio-demographic indices to capture the disparity between the wealthy and poor. Since their inception, they have also been used as a basis for modelling spread and disparity in other contexts. While recent research has identified that a number of classical inequality and welfare functions can be considered in the framework of OWA operators, here we propose a framework of penalty-based aggregation functions and their associated penalties as measures of inequality.
Abstract. The time needed to apply a hierarchical clustering algorithm is most often dominated by the number of computations of a pairwise dissimilarity measure. Such a constraint, for larger data sets, puts at a disadvantage the use of all the classical linkage criteria but the single linkage one. However, it is known that the single linkage clustering algorithm is very sensitive to outliers, produces highly skewed dendrograms, and therefore usually does not reflect the true underlying data structure - unless the clusters are well-separated.
To overcome its limitations, we proposed a new hierarchical clustering linkage criterion called genie. Namely, our algorithm links two clusters in such a way that a chosen economic inequity measure (e.g., the Gini or Bonferroni index) of the cluster sizes does not increase drastically above a given threshold.
Benchmarks indicate a high practical usefulness of the introduced method: it most often outperforms the Ward or average linkage in terms of the clustering quality while retaining the single linkage speed. The algorithm is easily parallelizable and thus may be run on multiple threads to speed up its execution further on. Its memory overhead is small: there is no need to precompute the complete distance matrix to perform the computations in order to obtain a desired clustering. In this talk we will discuss its reference implementation, included in the genie package for R.
Abstract. Since the 1980s, studies of aggregation functions most often focus on the construction and formal analysis of diverse ways to summarize numerical lists with elements in some real interval. Quite recently, we also observe an increasing interest in aggregation of and aggregation on generic partially ordered sets.
However, in many practical applications, we have no natural ordering of given data items. Thus, in this talk we review various aggregation methods in spaces equipped merely with a semimetric (distance). These include the concept of such penalty minimizers as the centroid, 1-median, 1-center, medoid, and their generalizations -- all leading to idempotent fusion functions. Special emphasis is placed on procedures to summarize vectors in Rd for d ≥ 2 (e.g., rows in numeric data frames) as well as character strings (e.g., DNA sequences), but of course the list of other interesting domains could go on forever (rankings, graphs, images, time series, and so on).
We discuss some of their formal properties, exact or approximate (if the underlying optimization task is hard) algorithms to compute them and their applications in clustering and classification tasks.
Abstract. The paper discusses a generalization of the nearest centroid hierarchical clustering algorithm. A first extension deals with the incorporation of generic distance-based penalty minimizers instead of the classical aggregation by means of centroids. Due to that the presented algorithm can be applied in spaces equipped with an arbitrary dissimilarity measure (images, DNA sequences, etc.). Secondly, a correction preventing the formation of clusters of too highly unbalanced sizes is applied: just like in the recently introduced Genie approach, which extends the single linkage scheme, the new method averts a chosen inequity measure (e.g., the Gini-, de Vergottini-, or Bonferroni-index) of cluster sizes from raising above a predefined threshold. Numerous benchmarks indicate that the introduction of such a correction increases the quality of the resulting clusterings.