For simplicity and ease of implementation, PGIA partitions the search space only once. If we continue partitioning the space, we might see some more improvements.
In some earlier tests, splitting the space too many times caused performance to degrade, as there were simply too many partitions.
So we'll need to find some way of limiting how many times we recursively split. (We could just keep a counter and limit it, or maybe estimate how many potential solutions are in a partition, and only split above some threshold.)
For simplicity and ease of implementation, PGIA partitions the search space only once. If we continue partitioning the space, we might see some more improvements.
In some earlier tests, splitting the space too many times caused performance to degrade, as there were simply too many partitions.
So we'll need to find some way of limiting how many times we recursively split. (We could just keep a counter and limit it, or maybe estimate how many potential solutions are in a partition, and only split above some threshold.)