A pseudo-quantum annealing software NEC Vector Annealing for quickly solving combinatorial optimization problems is available at OCTOPUS.

Basic use

Vector Annealing can be used from Python by reading the environment setting file with the following command.

$ module load BasePy
$ module load annealing/4.0.0

Here's how to run the sample script that solves the Traveling Salesman Problem:

* The sample script only supports problems with "EDGE_WEIGHT_TYPE: EUC_2D". Also, do not use it for large problems with more than 100 cities, as the QUBO may become too large to generate.

Please create a job script in the directory where the problem file is located, and execute by batch job. Please specify "VA" as the queue name. An example job script is as follows.

#!/bin/bash

#PBS -q VA

#PBS --group=[group name]

#PBS -l elapstim_req=00:05:00

#PBS -b 1

#PBS -v OMP_NUM_THREADS=64

cd ${PBS_O_WORKDIR}

module load BasePy

module load annealing/4.0.0

python /opt/va/V4.0.0X/samples/TSP/python/TspLib.py /opt/va/V4.0.0X/samples/TSP/python/eil51.tsp

You can also download the problem file (e.g., eil51.tsp) from TSPLIB and place it in any directory.

For specific instructions on how to use the Vector Annealing API, please refer to the following manuals.

Reference

Vector Annealing x86 version user's guide (Japanese) rev2
Vector Annealing C++API (x86 version) user's guide (Japanese) rev2
Vector Annealing Python API x86 version reference guide (Japanese) rev2
Parameter Estimator user's guide rev6