During the topological optimization process we start removing the smaller edges, guessing they are the less relevant. If such assumption is wrong we can always go back.
However, if a certain edge do not contribute to the amplitude of any existing ket, this may remain with its initial random value. Such value may not be the smallest, but the edge is still irrelevant. It may be worth to look for such circumstance, in case it actually occurs, to speed up the optimization process.
During the topological optimization process we start removing the smaller edges, guessing they are the less relevant. If such assumption is wrong we can always go back.
However, if a certain edge do not contribute to the amplitude of any existing ket, this may remain with its initial random value. Such value may not be the smallest, but the edge is still irrelevant. It may be worth to look for such circumstance, in case it actually occurs, to speed up the optimization process.