Algorithm final round about hashcode HOT 2 CLOSED

We implemented various methods using different approaches and performance metrics, take a closer look at the place_* methods to see the details.
I can briefly describe the goals of each approach:

• place_routers_on_skeleton
  Computes the skeleton of the available space ("rooms") and places router on it. Precalculates the available routers and used budget.

• place_routers_on_skeleton_iterative
  Same as place_routers_on_skeleton, but places router iteratively on randomized positions on the skeleton.

• place_routers_randomized
  Places routers iteratively on uniform sampled positions, updates used cables by computing a mst or via bsf to the closest places cable.

• place_routers_randomized_by_score
  Same as place_routers_randomized, but chooses a position based on the approximated score of that position.

• place_routers_by_convolution
  Computes scores of router position by convolving the available space with a constant squared matrix with size radius x radius. As a result of this, routers can 'see' through walls, which does not comply with the problem statement.

from hashcode.

Thank you!

from hashcode.