University of Twente Student Theses
Efficient Online Learning with Mondrian Forest on a FPGA
Nowee, Stijn (2025) Efficient Online Learning with Mondrian Forest on a FPGA.
Full text not available from this repository.
| Full Text Status: | Access to this publication is restricted |
| Embargo date: | 13 June 2026 |
| Abstract: | Online learning algorithms are crucial for applications processing real-time data streams, but their computational demands such as high throughput often necessitate hardware acceleration. Field-Programmable Gate Arrays (FPGAs) offer significant potential for acceleration due to their inherent parallelism. This thesis presents the design, implementation, and evaluation of the Mondrian Forest online learning algorithm on an FPGA platform. Mondrian Forests, while effective for online tasks, pose challenges for hardware implementation due to dynamic tree structures and large node sizes. To address this, we developed an FPGA architecture targeting the AMD Versal xcvh1742 card, leveraging its High-Bandwidth Memory (HBM) and parallel processing capabilities. The architecture employs multiple processing units, a page-based memory system to manage tree nodes efficiently, and specific adaptations to the algorithm, such as top-down posterior updates, for hardware suitability. Performance was evaluated using simulation and analytical methods on synthetic and KDD datasets. Results demonstrate substantial speedups compared to an optimized CPU (Rust-based Light-river) implementation, achieving 42.33x faster execution on the synthetic dataset and 56.42x on the KDD dataset, while maintaining high classification accuracy. While the implementation was slower (4.90x) than a specialized Adaptive Random Forest (ARF) FPGA implementation on the KDD dataset, it showed slightly better accuracy. This work confirms that FPGAs provide a high-performance platform for significantly accelerating Mondrian Forests, making them viable for demanding real-time online learning applications. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Embedded Systems MSc (60331) |
| Link to this item: | https://purl.utwente.nl/essays/106434 |
| 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