A prototype of the new solver, HOBO solver "hobotan," is now available.

We have created a new solver with volunteers on Discord.

We are extending QUBO to HOBO.

We are working on integrating it into the QUBO annealing compatible library TYTAN.

A prototype is available at https://github.com/ShoyaYasuda/hobotan.

You can try it with limited functionality, so please give us your feedback.

pip install -U git+https://github.com/ShoyaYasuda/hobotan

import numpy as np

from hobotan import *

"""
Only supports 1-dimensional arrays; [4, 4] is not allowed.
Only supports symbols named 'q'.
"""
q = symbols_list(5, 'q{}')

print(q)

"""
Expressions are probably OK with any number of terms.
"""
H = (q[0] + q[1] + q[2] + q[3] + q[4] - 5)**2 - 20*q[1]*q[2] - 10*q[1]*q[2]*q[3] + 5*q[1]*q[2]*q[3]*q[4]

Compile to HOBO tensor

hobo, offset = Compile(H).get_hobo()
print(offset)

print(hobo)

Select sampler (random seed fixed)

solver = sampler.SASampler(seed=0)

Sampling (100 times)

result = solver.run(hobo, shots=100)

Results

for r in result:
  print(r)
  arr, subs = Auto_array(r[0]).get_ndarray('q{}')
  print(arr)