Abstract: In the thesis a method is developed to find local clusters in a higher-dimensional data structure, i.e. a tensor, utilizing hierarchical clustering on the components of a tensor factorisation (namely the CP decomposition) under sparsity assumptions. The method was then validated and finally applied to real world gene expression data, where the data was collected over multiple cell lines and multiple treatments, therefore resulting in a tensor dataset. The method manages to uncover interesting structure in the tensor and shows robustness to noise.
Handledare: Felix Held
Examinator: Marina Axelson-Fisk