Setup

• Install poetry (pipx install poetry)
• Run poetry install --with data_explore

How to run an experiment

• Edit the config in config.yaml
• python graddft_qnn/main.py
• The current and past run results can be seen in reports.json
  • To assist merging different reports, we have consolidate_report.py
• Plot the train/test losses: train_test_qubits_plot.py

Current investigations

• The training loss is the same but the test is different
• Try to run on bigger number of qubits (maximum 9 qubits)

Commit rules

Please run ruff format . and ruff check --fix .

Datasets

1. (in use) Different bond lengths of H2 in training set and test set
2. H2 + Li2 + LiH
3. Different molecules with default Pennlylane bond length

Contribution

• A way to automatically generate ansatz given a group, regardless of the dimension or number of qubits
  • Both in matrix form and quantum gates form
• Invariant quantum neural network applied to DFT
• Performance comparison with normal n