Following tests with this first implementation, I found -as expected- that extension constraints make the graph explode. Moreover, I also found that many of them could be condensed. For example, say we have the following tuples:
(1, 2)
(1, 3)
(1, 4)
(1, 5)
(1, 6)
(1, 7)
(1, 8)
(1, 9)
(1, 10)
When we think about this constraint, we think that the first variable can take the value 1 and the second variable can take an integer value in the range {2, 3,..., 10}. Using a machine-readable analogous version would make the graphs a lot smaller, which in turn would make learning easier.
Following tests with this first implementation, I found -as expected- that extension constraints make the graph explode. Moreover, I also found that many of them could be condensed. For example, say we have the following tuples:
When we think about this constraint, we think that the first variable can take the value 1 and the second variable can take an integer value in the range {2, 3,..., 10}. Using a machine-readable analogous version would make the graphs a lot smaller, which in turn would make learning easier.