QP-SBGD: Projected Stochastic Gradient Descent with Quantum Annealed Binary Gradients (BMVC 2024)

Maximilian Krahn^1,2, Michele Sasdelli³, Fengyi Yang³, Vladislav Golyanik⁴, Juho Kannala¹, Tat-Jun Chin³ and Tolga Birdal²

¹ Aalto University , ²Imperial College London, ³ Adelaide University, ⁴ Max Plank Institute for Informatics.

This is the official repository for the project "Projected Stochastic Gradient Descent with Quantum Annealed Binary Gradients". In this work, we train binary neural networks with a quantum annealer deployable optimiser. The preprint can be found at https://arxiv.org/abs/2310.15128 and the project page can be found here. The code can be executed with PyTorch and the D-Wave ocean sdk. To run the code without a quantum annealer one can use D-Wave neal instead of a QPU Sampler or look at the experiments with gurobi.

Getting Started

• The repository can be cloned with
  git clone https://github.com/mk2510/QPSBGD/
  
• We recommend the user to set up a conda environment with

conda create --name QPSBGD --file requirements.txt

• The D-Wave ocean sdk can be installed with
  python -m pip install dwave-ocean-sdk
  and following the steps from https://docs.ocean.dwavesys.com/en/stable/overview/install.html. In particular to get access to the solvers from D-Wave one has to create an account for D-Wave Leap (https://cloud.dwavesys.com/leap/login/?next=/leap/) and can then use the API Token (https://docs.ocean.dwavesys.com/en/stable/overview/sapi.html). We tested the code for dwave-ocean-sdk versions 4.2 and 6.3.

• Installing Gurobi is another prerequisite. Please refer to the homepage to acquire a license.

10 Layer MLP - Adult Dataset

To reproduce the experiments of the paper it is required to download the datasets:

• wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/a1a
• wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/a1a

place those files in the folder datafolder.

The experiments are then executed by running the 10layers.ipynb file.

2 Layer MLP DWave

To reproduce the experiments of the paper it is required to download the datasets:

• wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/a1a
• wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/a1a

place those files in the folder datafolder.

The experiments are then executed by running the 2layers.ipynb file.

License

Shield:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.