/oscar

Primary LanguageJupyter Notebook

OSCAR: cOmpressed Sensing based Cost lAndscape Reconstruction

Reconstructing landscapes of variational quantum algorithms (VQAs) by compressed sensing.

TODO: link our paper.

Use cases and their visualization in our paper are generated using cs_*.py and vis_*.ipynb.

Commands that calling cs_*.py to generate those use cases are recorded in record.md.

TODO for release

  • whether to pack up Google and IBM data, and tutorial to unzip
  • pack up figs/grid_search and figs/optimization by linking Tianyi's repo
  • zip figs/grid_search_recon (too big) and put somewhere; add tutorial to unzip
  • sparsity data (Table IV)
  • data for Fig. 12, n=20
  • LICENSE

Installation

Recommend: create an Anaconda environment and install from source.

conda create -n oscar python=3.9
conda activate oscar
TODO: requirement

Download data:

sh ./download_data.sh
git clone https://github.com/kunliu7/oscar
cd oscar
pip install -e .
pytest

P.S. pytest might takes several minutes.

Data

TODO: link with QAOA-Simulator.

Examples

  • cs_comp_miti.py: compare mitigated landscapes

  • cs_distributed.py: recon. distributed landscapes

  • cs_evaluate.py: compute recon. error for p=1 and p=2

  • cs_high_dim_vary_2d.py: compute recon. error for high-dim landscapes

  • cs_opt_on_recon_landscapes.py: optimize on recon. landscapes by interpolation

  • cs_second_optimize.py: second optimization proposed in paper

  • vis_OSCAR_save_queries.py: visualize #Queries saved by OSCAR

Citation