University of Twente Student Theses
Accelerated Spectral Clustering using Random Forest and the Locally Linear Landmark approach
Herreveld, Z. van (2024) Accelerated Spectral Clustering using Random Forest and the Locally Linear Landmark approach.
|
PDF
1MB |
| Abstract: | In the era of big data advanced clustering techniques are in big demand. A popular technique in the field of unsupervised learning is spectral clustering. This is a graph-based clustering method that is able to capture complex data structures by leveraging the spectral properties of the data. This involves analyzing the eigenvalues and eigenvectors of certain matrices. However, particularly due to this eigendecomposition, the complexity of spectral clustering is very high. This research explores the potential of enhancing the complexity of spectral clustering by using extremely randomized trees to construct the similarity matrix and by implementing a dimensionality reduction technique called the Locally Linear Landmarks technique. By replacing binary trees with extremely randomized trees, significant reductions in runtime are achieved without losing performance. Additionally, our implementation of the LLL method yielded notable improvements in runtime, however with a notable trade-off in terms of performance in some cases. Lastly, this research contributes a successful Python implementation of the general spectral clustering algorithm. This implementation incorporates a random forest-based similarity measure, RatioRF, for calculating the similarity matrix and offers the option to use the Linear Locally Landmarks approach, paving the way for further research and advancements in the field of spectral clustering. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 31 mathematics, 54 computer science |
| Programme: | Applied Mathematics BSc (56965) |
| Link to this item: | https://purl.utwente.nl/essays/102225 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page