Table 2 We consider four problems in this paper, to which we refer by problem number or its tuple (N, n, α), where N is the total number of spins in the system, n is the number of fixed spins, and α is the number of auxiliary spins

From: Solving Boltzmann optimization problems with deep learning

Problem

Number of spins

Fixed spins

Auxiliary spins

Number of desired states

Dynamic range of h, J

 

N

n

α

  

1

9

4

1

16

[−4, 4]

2

11

5

1

32

[−64, 64]

3

14

6

2

128

[−256, 256]

4

15

6

3

192

[−256, 256]

  1. Problem 1 represents our 2-bit by 2-bit multiplication circuit configuration. Similarly, Problems 2, 3, and 4 represent our 2-bit by 3-bit, 2-bit by 4-bit, and 3-bit by 3-bit multiplication circuit configurations, respectively. The dynamic range column describes the continuous set of values that Hamiltonian coefficients h, J, that is the elements of ψ, can take in the SLSQP solver. The magnitude of the h, J values has an effect on the range of the output probabilities returned by Eq. (10). Values in this column were determined empirically to yield training data covering a wide range of output probabilities.
Back to article page