Date: November 2021
Person responsible for data: Adam Glos (aglos [at] iitis.pl).
The scripts necessary for generating the results provided in the "Space-efficient binary optimization for variational computing" preprint. We follow the notation introduced in the preprint.
- Julia v1.2.0
- Optim v0.21.0
- NPZ
- JLD2
- FileIo
- PyPlot
- Statistics
- LineSearches
- TravelingSalesmanExact
- GLPK
- other modules installed by default
The quantum evolution was emulated. Details of how the expected energy and gradient were calculated can be found in qaoa_utils.jl.
For optimization we used L-BFGS algorithm. Details on the optimization can be found in qaoa_optimizers.jl.
Hamilton problem:
julia hamilton_generator.jl hamilton 5
TSP problem:
The command will create the directory hamilton and generate Hamiltonians for 3, 4, 5 cities. For both HOBO and QUBO it assumes that the first city is visited at the first time-point. Thus for QUBO it requires (N-1)^2 qubits, and for HOBO (N-1) ceil(\log(N))qubits.
julia tsp_generator.jl tsp_3 3 100
julia tsp_generator.jl tsp_4 4 100
julia tsp_generator.jl tsp_5 5 100
The first command will create the directory tsp_3 and generate 100 TSP instances for 3 cities, similar for the rest of lines. For both HOBO and QUBO it assumes that the first city is visited at the first time-point. Thus for QUBO it requires (N-1)^2 qubits, and for HOBO (N-1)\lceil \log(N)\rceil qubits.
Before running experiments, run julia and type the following lines
julia> include("sparse_generator_loader.jl")
julia> generator(4)
julia> generator(6)
julia> generator(9)
julia> generator(12)
julia> generator(16)
These command generate necessary files for energy computation.
Experiment results for Hamilton problem can be generated using the following command
julia hamilton_qaoa_experiment.jl 24 100 15
where 24 is the number of cores used, 100 is the number of experiments made, and 15 is the maximal level considered.
Experiment results for TSP problem can be generated using following commands
julia tsp_qaoa_experiment.jl tsp_3 100 15 40 24
julia tsp_qaoa_experiment.jl tsp_4 100 15 40 24
julia tsp_qaoa_experiment-splitted.jl tsp_5 100 10 40 24 1
julia tsp_qaoa_experiment-splitted.jl tsp_5 100 10 40 24 2
julia tsp_qaoa_experiment-splitted.jl tsp_5 100 10 40 24 3
julia tsp_qaoa_experiment-splitted.jl tsp_5 100 10 40 24 4
where tsp_[n] is the directory with TSP instances, 100 is the number of TSP instances, 15 is the maximal level, 40 is the required number of successful optimization runs (trajectories) for each instance and each level, 24 is the number of cores used. The last number for tsp_5 cases indicates which input data are considered within the run, which enables simpler parallelization on PL-GRID infrastructure.
Plots were generated by the following commands
julia plotter.jl generate plot