Figure 3

AutoSOME benchmarking. AutoSOME performance is compared to common clustering methods by clustering seven diverse benchmark datasets (see Additional file 1 for datasets and details of the entire benchmarking comparison; AutoSOME method with circular node topology is shown here). All clustering methods, apart from AutoSOME, were provided with the benchmark number of clusters for each dataset. Boxplots (created in R [56] using default boxplot parameters) represent the distribution of F-measure scores for all benchmark datasets. An F-measure score of 1 represents a perfect solution. ('+' denotes mean F-measure, and vertical bars denote median F-measure; HC = Hierarchical Clustering; PCA = Principal Components Analysis).