SC²S Colloquium - June 20, 2018
|Date:||June 20, 2018|
|Time:||15:00 - 16:00|
Philipp Zetterer: Comparison of Different Cluster Analysis Algorithms Using Radio Measurement Data of Public Mobile Networks
This Master Thesis deals with Cluster analysis on mobile Network data. This data is normally used by mobile devices to communicate with the provider and their radio masts but can also be used to analyze the network. This data can have up to a few hundred variables which contain numbers as well as strings and missing values. Therefore selecting the feature space is one major task. Another one is to find configurations(of radio masts) that are used by clustering them and finding common values. In order to do that different approaches will be tested including K-Means, density based algorithms and others like Sparse Grids or statistical methods.
Keywords: Clustering, Mobile Networks, real-world data
Shreyas Shenoy: Towards Non-blocking Combination Schemes in the Sparse Grid Combination Technique
This is an Master's Thesis introduction advised by Michael Obersteiner
In this talk we will show you the first results for sparse grid communication technique, the issues faced during the implementation of the same and the direction will be heading towards in the future.
Keywords: Sparse grids, Non-Blocking
This is a Bachelor's Thesis submission talk advised by Thomas Huckle. Talk will be given in German.
As part of video compression, extensions of the matrix SVD compression are used to calculate the SVD compression of three-dimensional tensors in this thesis. Therefore, various tensors operations and methods are used to compress the video data. In the wake of this work, three compression strategies are tested with regards to various parameters. The test results show that the SVD video compression benefits from the applied fast Fourier transform and is suited especially in the field of grayscale video compression. Additionally, three different variants of colour video compression, three different data types and four compression ratios are tested and evaluated, whereby the separating colour variant and the 32-single data type deliver the best video compression quality. In the context of VP9, it is determined that the SVD video compression needs to be combined with other strategies or rather that it needs to be improved and adjusted in order to be able to apply it in modern compression standards.
Keywords: Tensor Methods, Compression, SVD