'likelihood': 'Binomial_' + str(nb_meas),
'normalize_Y': False
})
optim_b_max_step = copy.copy(optim_b)
optim_b_max_step['constraints'] = 'step_0.4_0_1'
optim_b_max_smooth = copy.copy(optim_b)
optim_b_max_smooth['constraints'] = 'smoothlin_0.1_0_1'
# ==================================================================
# Without constraints
# with binomial observations ()
#
#===================================================================
if (do_base):
# perfect measurement
optim = Learner.learner_Opt(model=model_base, **optim_base)
res_base = optim(track_learning=True)
p_base = res_base['params']
p_base_exp = res_base['params_exp']
func_base.theta = p_base
func_base.plot_function(x_plot)
model_base(p_base)
if (do_proj10):
# binomial measurement, gaussian likelihood
optim = Learner.learner_Opt(model=model_proj10, **optim_base)
res_proj10 = optim(track_learning=True)
p_proj10 = res_proj10['params']
p_proj10_exp = res_proj10['params_exp']
func_base.theta = p_proj10
func_base.plot_function(x_plot)
}
dico_simul = learner1DBH._process_controler(dico_simul)
dico_simul['control_obj'] = learner1DBH._build_control_from_string(
dico_simul['control_obj'], None, context_dico=dico_simul)
model = bh1d.BH1D(**dico_simul)
if (optim_type == 'BO2'):
#BO
optim_args = {
'algo': 'BO2',
'maxiter': 30,
'num_cores': 1,
'init_obj': 15,
'acq': 'EI'
}
optim = Learner.learner_Opt(model=model, **optim_args)
resBO2 = optim(track_learning=True)
resBO2['last_func'] = model.control_fun
print(resBO2)
res = resBO2
resBO2['opt_more']
if (optim_type == 'DE'):
optim_args = {'algo': 'DE', 'popsize': 5, 'maxiter': 75}
optim = Learner.learner_Opt(model=model, **optim_args)
resDE = optim()
print(resDE)
res = resDE
if (optim_type == 'BO'):
optim_args = {'algo': 'BO', 'maxiter': 250}
'state_init': 'GS_i',
'state_tgt': 'GS_inf',
'fom': fom,
'fom_print': True,
'track_learning': True,
'ctl_shortcut': 'owbds01_pwl15'
}
dico_simul = learner1DBH._process_controler(dico_simul)
dico_simul['control_obj'] = learner1DBH._build_control_from_string(
dico_simul['control_obj'], None, context_dico=dico_simul)
model = bh1d.BH1D(**dico_simul)
try:
func_used = pFunc_base.pFunc_base.read_func_from_file("SFtoMI_0")
except:
optim = Learner.learner_Opt(model=model, **optim_args)
resBO2 = optim(track_learning=True)
resBO2['last_func'] = model.control_fun
res = resBO2
func_used = model.control_fun
func_used.theta = res['params']
if (save):
func_used.save_to_file("SFtoMI_0")
#Testing
fom_test = fom + ['f2t2', 'fluence', 'smooth', 'varN']
dico_test = copy.copy(dico_simul)
dico_test['fom'] = fom_test
dico_test['track_learning'] = False
model_test = bh1d.BH1D(**dico_test)
optim_params = func_used.theta
'state_init': 'GS_i',
'state_tgt': 'GS_inf',
'fom': fom_GS,
'fom_print': True,
'track_learning': True,
'ctl_shortcut': 'owbds01_pwl15',
'kblock': 0,
'pblock': 1
}
dico_GS = learner1DBH._process_controler(dico_GS)
dico_GS['control_obj'] = learner1DBH._build_control_from_string(
dico_GS['control_obj'], None, context_dico=dico_GS)
model_GS = bh1d.BH1D(**dico_GS)
optim_GS = Learner.learner_Opt(model=model_GS, **optim_args)
res_GS = optim_GS(track_learning=True)
model_GS.control_fun.plot_function(np.arange(-0.01, T + 0.01, 0.01))
state_tmp = model_GS.EvolutionPopAdiab(nb_ev=2)
model_GS.plot_pop_adiab(plot_gap=True)
#==============================================================================
# Try to reach ES at the end
#==============================================================================
fom_firstE = ['projSS:neg_fluence:0.0001_smooth:0.05']
dico_firstE = {
'L': 2,
'Nb': 2,
'mu': 0,
'T': T,
'dt': 0.01,
# is there a better way
#===================================================================
optim_with_noise = True
optim_ideal = True
optim_with_noise_custom = True
optim_args = {
'algo': 'BO2',
'maxiter': 50,
'num_cores': 4,
'init_obj': 25,
'acq': 'EI'
}
func_test = dico_no_noise['control_obj']
if (optim_with_noise):
optim = Learner.learner_Opt(model=model_ensemble, **optim_args)
res_ensemble = optim(track_learning=True)
params_noise = res_ensemble['params']
params_noise_exp = res_ensemble['params_exp']
if (optim_with_noise_custom):
optim = Learner.learner_Opt(model=model_ensemble_custom, **optim_args)
res_ensemble_custom = optim(track_learning=True)
params_noise_custom = res_ensemble_custom['params']
params_noise_custom_exp = res_ensemble_custom['params_exp']
if (optim_ideal):
optim = Learner.learner_Opt(model=model_no_noise, **optim_args)
res_no_noise = optim(track_learning=True)
params_no_noise = res_no_noise['params']