lift_order uses the non-aggregated LineGraph of the previous order to create the next. This means there are a lot of duplicate topological paths (at different time) saved during computations. Although we cannot reduce the Graph to only have the unique paths because they might continue into different directions depending on the time, we can save only the unique node sequences and a corresponding index that saves at which position each duplicate of each unique node sequence should be.
The idea was already implemented in #156 for temporal shortest path calculations that are not needed anymore.
lift_orderuses the non-aggregated LineGraph of the previous order to create the next. This means there are a lot of duplicate topological paths (at different time) saved during computations. Although we cannot reduce the Graph to only have the unique paths because they might continue into different directions depending on the time, we can save only the unique node sequences and a corresponding index that saves at which position each duplicate of each unique node sequence should be. The idea was already implemented in #156 for temporal shortest path calculations that are not needed anymore.