Code for Hamiltonian-Oriented Homotopy QAOA

This is a repository for the code used to calculate the numerical results presented in the article "Hamiltonian-Oriented Homotopy QAOA" .

Software installation

The code was used on Ubuntu OS 20.04. It is not guarantee it will work on other operating systems

Anaconda

Anaconda distribution can be downloaded from https://www.anaconda.com/products/individual.

For this project, we created a conda environment with the required python packages. Type the following command on the terminal at the root project folder to install it:

conda env create -f mc-qaoa.yml

To activate the environment:

conda activate mc-qaoa 

Julia

Julia can be downloaded from the official website https://julialang.org/. The version 1.6.2 was used.

In order to set up the environment, please go to the directory src/qaoa_efficiency_analysis. Then run julia and use

julia> ]
(@v1.6) activate .
(@v1.6) instantiate

This will install the required packages based on Manifest.toml file. The environment will be activated on its own when running scripts.

Reproducing data

Objective Hamiltonian Generation

To generate the objective Hamiltonians, run the following script on the scr directory:

python hamiltonian_generate.py

This script generates 100 samples for each node value. The value of nodes are ranged from 6 to 18 with increment step equal to 2. The graph objects are saved in compare_data/graph/ and the objective Hamiltonians in compare_data/obj/.

Comparing energy evaluation

We compare the true energy based on total Hamiltonian eigenvalues for 10 nodes running the following script:

python true_energy.py

This script calculates the energy for different initial $\alpha$ values and step. The data is saved on compare_data/data/.

To calculate the QAOA and HOHo-QAOA energies, run the the following commands on the julia scripts folder qaoa_efficiency_analysis:

for VAR in {2..18..2}; do julia sparse_generator_preparation.jl $VAR; done

This script prepares the sparse matrices used for the optimization. Then:

julia maxcut_qaoa_experiment.jl

To evaluate the data generated in true_energy.py and maxcut_qaoa_experiment.jl, run the script:

python plot_data_generate.py

And plot the results with:

python plot.py

QAOA energy landscape

We generate and plot data for energy landscape for QAOA with the following command:

python qaoa_energy_landscape_plot.py

QAOA methods comparison

To run the optimization experiments for QAOA, T-QAOA and HOHo-QAOA, use the following command:

julia maxcut_multi_qaoa_experiment.jl 

The data is saved by default in "scr/compare_data/data/<graph_type>/data_<date_of_experiment>. Alternatively, you can pass an argument for saving the data in the a directory giving its path:

julia maxcut_multi_qaoa_experiment.jl -outnew <path/to/directory>

If the a result in the directory already exists, it will be skipped.

For the paper, we saved the experiment esults on the following directories:

scr/compare_data/data/barabasi-albert/5_to_100_layers
scr/compare_data/data/barabasi-albert/t-vs-qaoa
scr/compare_data/data/barabasi-albert/t-zero-rand
scr/compare_data/data/barabasi-albert/5_to_100_layers

and the plots were generated with:

julia plotting_code.jl

The plots will be saved to scr/plot.