def solve_multiple_frontier(saving=False):
bbax=bax()
frontiers = {fn: Frontier(fn) for fn in SWB_functions}
best = {fn: None for fn in objective_functions}
indiv = [Person(theta[i],maxCoord = sizeRegion) for i in range(nIndiv)]
fac = [Facility(maxCoord = sizeRegion) for i in range(nFac)]
if loading:
dist = pickle.load(open(dumpFileName,'rb'))
else:
dist = np.array([[person.dist(facility) for facility in fac] for person in indiv])
total_subsets = bbax.enumerator.choose(nFac,nSelectedFac)
for ind, gp in enumerate(bbax.bax_gen(nFac,nSelectedFac)):
if ind % 1000 == 0:
num_in_each_frontier = [len(frontier.instances) for fn, frontier in frontiers.items()]
print(f"Processed {ind} subsets out of {total_subsets}")
print(num_in_each_frontier)
gp = gp.toList()
r = [min((dist[i][j] for j in gp)) for i in range(nIndiv)]
# print(len(r))
yv = np.zeros((nFac),dtype=int)
yv[gp] = 1
instance = Instance(rvals=r, yvals=yv)
for fn,frontier in frontiers.items():
frontier.consider(instance)
best = {fn : instance if instance.beats(best_instance,fn) else best_instance for fn,best_instance in best.items()}
# if saving:
# save_data(dat)
return best,frontiers,dist
def solve_enumerative(obj_fn):
bbax = bax()
yvals = np.zeros((nFac), dtype=int)
rvals = np.zeros((nIndiv))
uvals = np.full((nIndiv), -np.inf)
for ind, gp in enumerate(bbax.bax_gen(nFac, nSelectedFac)):
gp = gp.toList()
xlist = [
gp[np.argmin([dist[i][j] for j in gp])] for i in range(nIndiv)
]
r = [dist[i][xlist[i]] for i in range(nIndiv)]
u = obj_fn(r)
print(f"Obj is {sum(u)}")
if sum(u) > sum(uvals):
uvals = u
rvals = r
yvals = np.zeros((nFac), dtype=int)
yvals[gp] = 1
print(f"Best obj is {sum(uvals)}")
fstar = sum(uvals)
prob_success = [
1 / (1 + exp(-beta[0] - beta[1] * theta[i] - beta[2] * rvals[i]))
for i in range(nIndiv)
]
print(f"Solved enumeratively")
return yvals, rvals, uvals, prob_success, fstar, obj_fn.extra_info
def solve_multiple_frontier(saving=False):
bbax = bax()
frontiers = {
fn: [Frontier(fn) for trial in range(nTrials)]
for fn in SWB_functions
}
all_best = {fn: None for fn in objective_functions}
for trial in range(nTrials):
best = {fn: None for fn in objective_functions}
indiv = [Person(theta[i], maxCoord=sizeRegion) for i in range(nIndiv)]
fac = [Facility(maxCoord=sizeRegion) for i in range(nFac)]
dist = np.array([[person.dist(facility) for facility in fac]
for person in indiv])
for ind, gp in enumerate(bbax.bax_gen(nFac, nSelectedFac)):
gp = gp.toList()
r = [min((dist[i][j] for j in gp)) for i in range(nIndiv)]
# print(len(r))
yv = np.zeros((nFac), dtype=int)
yv[gp] = 1
instance = Instance(rvals=r, yvals=yv)
for fn in frontiers:
frontier = frontiers[fn][trial]
if not frontier.dominates(instance):
frontier.append(instance)
# for fn, best_instance in best.items():
# if instance.beats(best_instance,fn):
# best[fn] = instance
best = {
fn: instance
if instance.beats(best_instance, fn) else best_instance
for fn, best_instance in best.items()
}
for fn, combined_instance in all_best.items():
all_best[fn] = best[fn].combine(combined_instance)
print(f"Solved {trial+1}/{nTrials} enumeratively")
# if saving:
# save_data(dat)
return all_best, frontiers
def solve_multiple_enumerative(saving=False):
bbax = bax()
all_yvals = {fn: [] for fn in objective_functions}
all_rvals = {fn: [] for fn in objective_functions}
all_uvals = {fn: [] for fn in objective_functions}
all_fstar = {fn: [] for fn in objective_functions}
all_prob_success_vals = {fn: [] for fn in objective_functions}
for trial in range(nTrials):
indiv = np.random.uniform(0, sizeRegion, (nIndiv, 2))
fac = np.random.uniform(0, sizeRegion, (nFac, 2))
dist = np.array([[
(indiv[i, 0] - fac[j, 0])**2 + (indiv[i, 1] - fac[j, 1])**2
for j in range(0, nFac)
] for i in range(0, nIndiv)]) # nIndiv X nFac
yvals = {fn: np.zeros((nFac), dtype=int) for fn in objective_functions}
rvals = {fn: np.zeros((nFac), dtype=int) for fn in objective_functions}
uvals = {fn: np.full((nIndiv), -np.inf) for fn in objective_functions}
fstar = {fn: 0 for fn in objective_functions}
prob_success_vals = {fn: 0 for fn in objective_functions}
for ind, gp in enumerate(bbax.bax_gen(nFac, nSelectedFac)):
gp = gp.toList()
# print(locals())
# print(f"gp: {gp}, gp_dists: {[dist[i][j] for j in gp]}, ")
xlist = [
gp[np.argmin([dist[i][j] for j in gp])] for i in range(nIndiv)
]
r = [dist[i][xlist[i]] for i in range(nIndiv)]
for f in objective_functions:
u = f(r)
# u = [-np.log(1+np.exp(-beta[0] - beta[1]*theta[i] - beta[2]*r[i])) for i in range(nIndiv)]
if sum(u) > sum(uvals[f]):
uvals[f] = u
rvals[f] = r
yvals[f] = np.zeros((nFac), dtype=int)
yvals[f][gp] = 1
for f in objective_functions:
fstar[f] = sum(uvals[f])
prob_success_vals[f] = prob_success(rvals[f], beta, theta)
all_yvals[f].extend(yvals[f])
all_rvals[f].extend(rvals[f])
all_uvals[f].extend(uvals[f])
all_fstar[f].append(fstar[f])
all_prob_success_vals[f].extend(prob_success_vals[f])
print(f"Solved {trial+1}/{nTrials} enumeratively")
for f in objective_functions:
all_yvals[f] = tuple(
all_yvals[f]) # so they are hashable later for grouping plots
dat = {
"yvals": all_yvals,
"rvals": all_rvals,
"uvals": all_uvals,
"fstar": all_fstar,
"prob_success_vals": all_prob_success_vals
}
if saving:
save_data(dat)
return dat
