TreeLUT: An Efficient Alternative to Deep Neural Networks for Inference Acceleration Using Gradient Boosted Decision Trees Conference: Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays; February 27-March 1, 2025; Monterey, CA, USA Proceedings of the 2025 ACM
Alireza Khataei, Kia Bazargan · 2025
Read TreeLUT: An Efficient Alternative to Deep Neural Networks for Inference Acceleration Using Gradient Boosted Decision Trees Conference: Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays; February 27-March 1, 2025; Monterey, CA, USA Proceedings of the 2025 ACM by Alireza Khataei, Kia Bazargan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-01511.txt.
Description
Alireza Khataei and Kia Bazargan wrote this paper in two thousand twenty-five. It presents TreeLUT, an open tool that runs gradient boosted decision trees on FPGA chips through lookup tables, as a lean rival to neural nets.
Neural nets lean on vast counts of multiply and add steps, which cost much on hardware. Boosted trees, built from plain decision trees, match the shape FPGAs handle best, and their run steps are mere compares and adds. TreeLUT adds a quantization scheme and a pipeline design, using lookup tables alone with no block memory or DSP units. On standard test sets it beats a row of neural and tree rivals on chip use, delay, and throughput, at matched accuracy.
Catalog record
| CREATOR | Alireza Khataei, Kia Bazargan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-01511