def loop(a, level, parass, results, results_global, vers=version):
"""
New in 2.6: Makes an arbitrary number of loops
a: dictionnary, with keys as parameters to loop on and values as the values on which to loop.
level: list of parameters on which to loop. The first one is the most outer loop, the last one is the most inner loop.
"""
if level==[]:
#rep, name= build_path_average(paras, vers=vers)
# print parass['par']
# print parass['AC_dict']
rep, name= build_path_average(parass, vers=vers)
with open(rep + name, 'r') as f:
results=pickle.load(f)
# print 'results'
# print results
# print
#results_global={met:{k:sum([results[met][p][k]*paras['AC_dict'][p]/float(paras['ACtot']) \
# for p in results[met].keys()]) for k in ['avg', 'std']} for met in ['satisfaction', 'regulated_FPs', 'regulated_F']}
results_global={met:{k:sum([results[met][p][k]*parass['AC_dict'][p]/float(parass['ACtot']) \
for p in results[met].keys() if parass['AC_dict'].has_key(p)]) for k in ['avg', 'std']} for met in ['satisfaction', 'regulated_FPs', 'regulated_F']}
return results,results_global
else:
assert level[0] in a.keys()
for i in a[level[0]]:
# print level[0], '=', i
parass.update(level[0],i)
# print parass['AC_dict']
results[i], results_global[i]=loop(a, level[1:], parass, {}, {}, vers=vers)
return results, results_global
def average_sim(paras=None, G=None, save=1, suffix='', stat_distance='common_area'):
"""
New in 2.6: makes a certain number of iterations (given in paras) and extract the averaged mettrics.
Change in 2.7: parallelized.
Changed in 2.9.1: added force
"""
def do((paras, G,i)):
rlts={}
sim=Simulation(paras, G=G.copy(), verbose=False)
sim.make_simu(storymode=False)
sim.queue=post_process_queue(sim.queue)
rlts = compare_networks(sim, stat_distance = stat_distance)
del sim
return rlts
rep, name = build_path_average(paras, Gname=paras['G'].name)
name = name[:-4] + suffix + '.pic' #_comparison_distributions.pic'
if paras['force'] or not os.path.exists(rep + name):
inputs = [(paras, G, i) for i in range(paras['n_iter'])]
start_time=time()
if paras['parallel']:
print 'Doing iterations',
results_list = parmap(do, inputs)
else:
results_list=[]
for i,a in enumerate(inputs):
sys.stdout.write('\r' + 'Doing simulations...' + str(int(100*(i+1)/float(paras['n_iter']))) + '%')
sys.stdout.flush()
results_list.append(do(a))
print '... done in', time()-start_time, 's'
results={}
for met in results_list[0].keys():
#print met, results_list[0][met], type(results_list[0][met]), type(np.float64(1.0))
if type(results_list[0][met])==type(np.float64(1.0)) or type(results_list[0][met])==type(1.0):
results[met]={'avg':np.mean([v[met] for v in results_list]), 'std':np.std([v[met] for v in results_list])}
elif type(results_list[0][met])==type({}):
results[met]={tuple(p):[] for p in results_list[0][met].keys()}
for company in results_list[0][met].keys():
results[met][company]={'avg':np.mean([v[met][company] for v in results_list]), 'std':np.std([v[met][company] for v in results_list])}
if save>0:
#rep, name=build_path_average(paras, Gname=G.name)
os.system('mkdir -p ' + rep)
with open(rep + name,'w') as f:
pickle.dump(results, f)
#print 'Saving in', rep + name
else:
print 'Skipped this value because the file already exists and parameter force is deactivated.'
print 'I load it from disk.'
with open(rep + name,'r') as f:
results = pickle.load(f)
return results
def build_path(paras, vers='2.9', **kwargs):
rep, name = build_path_average(paras, vers = vers, **kwargs)
return rep, name + '_shocks'
