@inproceedings{tang_dt2hdl_2025,
	title = {{DT}2HDL: A Binary Decision Tree to {HDL} Generation Tool},
	issn = {1948-3295},
	url = {https://ieeexplore.ieee.org/document/11014334},
	doi = {10.1109/ISQED65160.2025.11014334},
	shorttitle = {{DT}2HDL},
	abstract = {Edge computing devices ideally need to process large amounts of sensor data and execute machine learning models in real-time. Deploying deep learning algorithms can be challenging in energy-constrained nodes with limited computational capacity, when using lightweight classification algorithms is essential. Decision Tree ({DT}) algorithms require fewer computational resources than Artificial Neural Networks ({ANNs}) while maintaining similar accuracy. However, implementing {DTs} on hardware becomes difficult as the number of nodes increases. This work proposes a framework for automatically generating Hardware Description Language ({HDL}) code for binary decision trees, simplifying their deployment on {FPGAs} or {ASICs}. The effectiveness of the {DT}2HDL tool was evaluated using five distinct datasets: electrical fault detection, electrical fault classification, heart disease, breast cancer, and Iris. Results from both {HDL} and Python simulations indicate that full-precision decision trees ({DTs}) and 8-bit quantized {DTs} achieve the same accuracy levels as artificial neural networks ({ANNs}) with more than five hidden neurons.},
	eventtitle = {2025 26th International Symposium on Quality Electronic Design ({ISQED})},
	pages = {1--8},
	booktitle = {2025 26th International Symposium on Quality Electronic Design ({ISQED})},
	author = {Tang, Qijia and Pamunuwa, Dinesh and Weerasekera, Roshan},
	urldate = {2025-12-15},
	date = {2025-04},
	keywords = {{ASIC}, Accuracy, Artificial neural networks, Classification algorithms, Codes, Decision Tree, Decision trees, Edge Computing, Edge computing, Electrical Fault, {FPGAs}, Field programmable gate arrays, Hardware design languages, Internet of Things, Machine learning algorithms, Neural Network, Support vector machines, publication},
}