##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
Removing the k dimension
Add penalty per region. Penalized DP: = min cost to partition first levels into any number of regions, with penalty. Transition: . The is the per-region penalty.
Now you have states (one per position). If can be computed in with preprocessing, total time is per binary search iteration, or overall.
