Two-way incremental seriation in the temporal domain with three-dimensional visualization: Making sense of evolving high-dimensional datasets

Abstract

Two-way seriation is a popular technique to analyse groups of similar instances and their features, as well as the connections between the groups themselves. The two-way seriated data may be visualized as a two-dimensional heat map or as a three-dimensional landscape where colour codes or height correspond to the values in the matrix. To achieve a meaningful visualization of high-dimensional data, a compactly supported convolution kernel is introduced, which is similar to filter kernels used in image reconstruction and geostatistics. This filter populates the high-dimensional space with values that interpolate nearby elements, and provides insight into the clustering structure. Ordinary two-way seriation is also extended to deal with updates of both the row and column spaces. Combined with the convolution kernel, a three-dimensional visualization of dynamics is demonstrated on two data sets, a news collection and a set of microarray measurements.