University of Twente Student Theses
Comparing random forest implementations, BDT against MDD from BDT
Wolde, Jop ten (2024) Comparing random forest implementations, BDT against MDD from BDT.
|
PDF
444kB |
| Abstract: | Multi-valued decision diagrams (MDDs) show a promising path to potentially higher evaluating speed of random forest. Only the Treelite library and by extension the TL2Cgen library have no support for these models. They do however support binary decision trees (BDTs). This paper is an exploration of the conversion of BDTs to MDDs. The main advantage of an MDD is that each feature only needs to be evaluated in one node. Treelite however does not support nodes with more than two edges, which would be needed to create the MDD nodes. They were instead emulated using binary nodes. Comparing the MDD and BDT shows that the MDD is equal to the BDT at best in this implementation. The closest results were achieved by having only a few features per tree. Trees with more features resulted in an increase in tree depth and a decrease in prediction per second. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Programme: | Electrical Engineering BSc (56953) |
| Link to this item: | https://purl.utwente.nl/essays/104813 |
| 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