Tensor Clusters for Extracting and Summarizing Components in Spectral Cubes

Modern Integral Field Spectrograph instruments like the Multi Unit Spectroscopic Explorer generate large-scale data. The data size problem has been studied in recent times but there is another equally relevant problem, the data cubes also exhibit unprecedented complexity and dimensionality. These complex sources can be modeled using tensor algebraic methods. In this representation the cube has two physical axes (Ra/Dec), and a 3rd axis (Wavelength, Time, Intensity). The representation is a tensor of order 3 that leads to a concise description astronomical data cubes. We applied over this representation a tensor-clusters approach to find the morphology of components expected in these complex data cubes. To achieve this goal the TensorFit, a library with strong GPU support to handle spectral cubes in a tensor mode, was used.