def random_search_2(solution):
""" Random search to a solution. More informed than random_search.
While not solved, the method chooses an unsatisfied vertex, with
probability proportional to its demand. It then clears a route
from the supply center to that vertex, and repeats.
"""
vs = solution.vertices
es = solution.edges
while not solution.satisfied():
#pick an unsatisfied node, and go there
todo = list(solution.unsatisfied_vertices)
pri = [t.demand for t in todo]
if len(todo) != 0:
target = choice_prob(todo, pri)
source = choice(solution.supplies)
[e.clear() for e in solution.decisions]
path = min_path(vs, es, source, target)
[e.restore() for e in solution.decisions]
path = filter(lambda x: x not in solution.decisions, path)
solution = solution.step_many(path)
else:
print 'defaulting to greedy'
return greedy_search(solution)
return solution
from network import min_path
from random import choice
v,e,r = read('cambridge')
supplies = [vv for vv in v if vv.supply > 0]
s = Solution(v, e, r, supplies)
s = greedy_search(s)
s.plot(out='nudge1.png')
d = s.decisions
truncate = 40
begin = d[:truncate]
satisfied = set([vv for ee in begin for vv in [ee.v1, ee.v2]])
unsatisfied = list(set(v).difference(satisfied))
target = choice(unsatisfied)
[ee.clear() for ee in begin]
middle = min_path(v, e, supplies[1], target)
[ee.restore() for ee in begin]
end = d[truncate:]
s2 = Solution(v, e, r, supplies, decisions = begin)
s2.plot(out='nudge2.png')
s3 = Solution(v, e, r, supplies, decisions = begin + middle)
s3.plot(out='nudge3.png')
s4 = Solution(v, e, r, supplies, decisions = begin + middle + end)
s4.plot(out='nudge4.png')
