OpenHD

OpenHD - A GPU-Powered Framework for Hyperdimensional Computing

• Author: Jaeyoung Kang (UCSD) and Yeseong Kim (DGIST)

The OpenHD framework enables the GPU-based execution of HD Computing using JIT-like compliation written in Python fo high efficiency. For the implementation details, please refer to our papers in the references section below.

Requirements

We included the library dependencies in the pip installer. You also need to install GraphViz for debuging purpose.

# apt install graphviz

Install

You can install the OpenHD framework using pip3:

pip install .

Usage

An usage example is provided in example/voicehd.py:

python3 examples/voicehd.py -t examples/dataset/isolet_train.choir_dat -i examples/dataset/isolet_test.choir_dat

References

@article{kang2022openhd,
  title={OpenHD: A GPU-Powered Framework for Hyperdimensional Computing},
  author={Kang, Jaeyoung and Khaleghi, Behnam and Rosing, Tajana and Kim, Yeseong},
  journal={IEEE Transactions on Computers},
  year={2022},
  publisher={IEEE}
}

@inproceedings{kang2022xcelhd,
  title={XCelHD: An efficient GPU-powered hyperdimensional computing with parallelized training},
  author={Kang, Jaeyoung and Khaleghi, Behnam and Kim, Yeseong and Rosing, Tajana},
  booktitle={2022 27th Asia and South Pacific Design Automation Conference (ASP-DAC)},
  pages={220--225},
  year={2022},
  organization={IEEE}
}