. 2020 May 12;3:33. doi: 10.3389/frai.2020.00033

Algorithm 2.

Targeting a balanced dataset

Input: Target rate PR_t, biased acceptance rate

{P R}_{b}^{i}

Output: Transition probabilities

d_{ȳ = j}^{s = i}

1: if

{P R}_{t} > {P R}_{b}^{i}

then

2: ℙ(ȳ = 1|y = 1, s = i)←1

3: else

ℙ (ȳ = 1 | y = 1, s = i) \leftarrow \frac{{P R}_{t}}{{P R}_{b}^{i}}

5: end if

6: if j=0 then

7: ℙ(ȳ = 0|y = 1, s = i)←1−ℙ(ȳ = 1|y = 1, s = i)

d_{ȳ = 0}^{s = i} \leftarrow \frac{ℙ (ȳ = 0 | y = 1, s = i) \cdot {P R}_{b}^{i}}{1 - {P R}_{t}}

9: else if j=1 then

10:

d_{ȳ = 1}^{s = i} \leftarrow \frac{ℙ (ȳ = 1 | y = 1, s = i) \cdot {P R}_{b}^{i}}{{P R}_{t}}

11: end if