10.18453/rosdok_id00005109 Beese, Martina Martina Beese http://d-nb.info/gnd/138990959X https://orcid.org/0009-0001-2849-5960 Universität Rostock 2024 510 Mathematics 2024 en https://purl.uni-rostock.de/rosdok/id00005109 urn:nbn:de:gbv:28-rosdok_id00005109-5 The nonnegative matrix factorization (NMF) problem does not have a unique solution. A low-dimensional representation can be used to determine the set of solutions. However, the matrix to be factorized may be incomplete, making the common approaches to determine the set of solutions only applicable to the largest complete submatrix. This thesis shows a way to approximate the set of solutions of the NMF problem for incomplete matrices with maximum utilization of the given information. This is done using approaches from cone theory, which allow to represent the given incomplete matrix.