The reason why I considering the node where the score is at least the same with the best score is due to the need to expand the region to at least the number of qubits that we required
Why this works: as the filtering now limit the region to what we want, the node that we assuming should satisfy the filtering and bounded by the region that we want. Therefore, by using union to take into account all regions, we hope to cover the amount of qubits that we want.
After the pair coding session: It seems that the surround function has the bug of not considering the branches that we expect as those points have been examined and had the same score with the best score so far.
The reason why I considering the node where the score is at least the same with the best score is due to the need to expand the region to at least the number of qubits that we required Why this works: as the filtering now limit the region to what we want, the node that we assuming should satisfy the filtering and bounded by the region that we want. Therefore, by using union to take into account all regions, we hope to cover the amount of qubits that we want.
After the pair coding session: It seems that the surround function has the bug of not considering the branches that we expect as those points have been examined and had the same score with the best score so far.