def resolve_sub_problem_lagrangean(costs, incidence_matrix,columns, dual_lagrangean):
set_covering = SetCoveringLagrangeanProblem(costs=costs, incidence_matrix=incidence_matrix,columns=columns, dual_lagrangean=dual_lagrangean)
set_covering.create_cplex_problem()
set_covering.c.solve()
x = set_covering.c.solution.get_values(set_covering.names_variables)
objective = set_covering.c.solution.get_objective_value()
return objective, x
def main(name_file, UB):
MAX_ITER = 500000
MAX_TIME_SEC = 7200
costs, incidence_matrix = my_io.read_file_format_or_library(name_file)
columns = transform_incidence2column(incidence_matrix)
#passo0
dual_lagrangean = [1]*len(incidence_matrix)
start = colect_time.cpu_time()
set_covering = SetCoveringLagrangeanProblem(costs=costs, incidence_matrix=incidence_matrix,columns=columns, dual_lagrangean=dual_lagrangean)
set_covering.create_cplex_problem()
set_covering.c.solve()
m1 = 0.001
x = set_covering.c.solution.get_values(set_covering.names_variables)
v = subtract_two_vectors([1]*len(incidence_matrix), multiply_matrix_by_vector(incidence_matrix, x))
z = x[:]
dual_lagrangean_aux = dual_lagrangean[:]
w = v[:]
p = dual_lagrangean[:]
epsilon = 0
t = 1
k = 1
s = 0.1
objective = 0
while True:
#passo1
dual_lagrangean = add_two_vectors(dual_lagrangean_aux, number_multiply_vector(s, w))
ro = s * product_dot(w, w) + abs(product_dot(w, subtract_two_vectors(dual_lagrangean_aux,p))) + epsilon
if stop(w,dual_lagrangean_aux,p) or UB-objective < 1 or colect_time.cpu_time() - start > MAX_TIME_SEC or t > MAX_ITER:
end = colect_time.cpu_time()
break
#passo2:
objective, x = resolve_sub_problem_lagrangean(costs=costs, incidence_matrix=incidence_matrix,columns=columns, dual_lagrangean=dual_lagrangean)
v = subtract_two_vectors([1]*len(incidence_matrix), multiply_matrix_by_vector(incidence_matrix, x))
#passo3
objective_aux, x_aux = resolve_sub_problem_lagrangean(costs=costs, incidence_matrix=incidence_matrix,columns=columns, dual_lagrangean=dual_lagrangean_aux)
if objective >= objective_aux + m1 * ro:
dual_lagrangean_aux = dual_lagrangean[:]
k = k + 1
#passo4
s = 0.5*((UB-objective)/product_dot(w,w))
alpha = resolve_problem_aux(v, w, s, product_dot(v, subtract_two_vectors(dual_lagrangean_aux,dual_lagrangean)), product_dot(w, subtract_two_vectors(dual_lagrangean_aux, p))+epsilon)
z = add_two_vectors(number_multiply_vector(alpha, x),number_multiply_vector(1-alpha, z))
w = add_two_vectors(number_multiply_vector(alpha, v),number_multiply_vector(1-alpha, w))
p = add_two_vectors(number_multiply_vector(alpha, dual_lagrangean),number_multiply_vector(1-alpha, p))
aux = 1-alpha * product_dot(subtract_two_vectors(v, w), subtract_two_vectors(p, dual_lagrangean))
epsilon = alpha * aux + (1 - alpha) * epsilon
t = t + 1
print name_file, objective, t, end