def test():
# Hypersurface shape and pygor
box_dim = 2
box_a = np.array([-1300., -1000.])
box_b = 500. * np.ones(box_dim)
shape = pygor_dummy.Box(ndim=box_dim, a=box_a, b=box_b)
th_leak = -500. * np.ones(2)
shape = pygor_dummy.Leakage(shape, th_leak)
origin = 0. * np.ones(box_dim)
# Dummy function for extra measurement
box_peaks = ([-1400., -1100], [-750, -900])
box_goodscore = ([-1400., -1100], [-1100, -900])
do_extra_meas = DummyExtMeas(box_peaks, box_goodscore)
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor_dummy(
shape)
# Poff detector
step_back = 100 # important param for measuring d
len_after_pinchoff = 100
threshold_high = 0.8
threshold_low = 0.2
d_r = 10
detector_pinchoff = util.PinchoffDetectorThreshold(
threshold_low) # pichoff detector
detector_conducting = util.ConductingDetectorThreshold(
threshold_high) # reverse direction
tester = Tester(pg,
lb_short,
ub_short,
detector_pinchoff,
d_r=d_r,
len_after_pinchoff=len_after_pinchoff,
logging=True,
detector_conducting=detector_conducting,
set_big_jump=set_big_jump,
set_small_jump=set_small_jump)
# Initial observations
num_init = 10
u_all = random_hypersphere(box_dim, num_init)
u_all, r_all, d_all, poff_all, detected_all, time_all, extra_meas_all = rw.get_full_data(
u_all, tester, step_back, origin=origin)
# GP hypersurface model
###
# Gaussian process for r
###
num_active_gates = box_dim
l_prior_mean = 0.2 * np.ones(num_active_gates)
l_prior_var = 0.1 * 0.1 * np.ones(num_active_gates)
r_min, r_max = 0.0, np.sqrt(num_active_gates) * 2000.
v_prior_mean = ((r_max - r_min) / 4.0)**2
v_prior_var = v_prior_mean**2
noise_var_r = np.square(d_r / 2.0)
gp_r = GP_util.create_GP(num_active_gates, 'Matern52', v_prior_mean,
l_prior_mean, (r_max - r_min) / 2.0)
GP_util.set_GP_prior(gp_r, l_prior_mean, l_prior_var, None,
None) # do not set prior for kernel var
GP_util.fix_hyperparams(gp_r, False, True)
# GP with initial observations
#gp_r.create_model(u_all, r_all[:,np.newaxis], noise_var_r, noise_prior='fixed')
#gp_r.optimize(num_restarts=20, opt_messages=False, print_result=True)
#plot_example(lb_short, ub_short, shape, gp_r, u_all, r_all, origin, 'initial')
#print(np.array(sampler.history_all).shape) # (num_samples, num_iter, ndim)
#plot_example(lb_short, ub_short, shape, gp_r, u_all, r_all, origin, 'after')
num_samples = 100
do_random_meas(box_dim,
num_samples,
tester,
step_back,
origin,
lb_short,
ub_short,
gp_r,
do_extra_meas=do_extra_meas,
save_dir=None)
def main():
conf_name = 'dummy'
if conf_name == 'config1':
conf_func = config.config1
ip = "http://129.67.86.107:8000/RPC2"
save_dir = Path('./save_Dominic_ABL_group_optparam_run2')
settletime, settletime_big = 0.01, 0.03 # settletime_big is for shuttling, 'None' disables shuttling
# ['c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9', 'c10']
origin = -100.
threshold_low = 0.0 # for pinchoff detector
d_r = 10 # length of one step
elif conf_name == 'config2':
# ['c3', 'c4', 'c8', 'c10', 'c11', 'c12', 'c16']
conf_func = config.config2
ip = 'http://129.67.85.38:8000/RPC2'
save_dir = Path('./save_Basel2_group')
settletime, settletime_big = 0.02, 0.02 # settletime_big is for shuttling, 'None' disables shuttling
origin = -100.
threshold_low = 2.e-11
d_r = 10 # length of one step
elif conf_name == 'dummy':
import pygor_dummy
box_dim = 5
box_a = -1000. * np.ones(box_dim)
box_b = 1000. * np.ones(box_dim)
shape = pygor_dummy.Box(ndim=box_dim, a=box_a, b=box_b)
th_leak = np.array([-500., -400., -300., 0., 0.])
shape = pygor_dummy.Leakage(shape, th_leak)
save_dir = Path('./save_dummy')
origin = -100.
threshold_low = 0.2
d_r = 10 # length of one step
else:
raise ValueError('Unsupported setup')
save_dir.mkdir(exist_ok=True)
if conf_name != 'dummy':
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor(conf_func, ip, settletime, settletime_big)
else:
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor_dummy(shape)
threshold_high = 0.8 * max_current
num_active_gates = len(active_gate_idxs)
# gate names are for nothing
active_gate_names = ["c{}".format(i+1) for i in active_gate_idxs]
print(active_gate_names)
''' Disable grouped gates
new_wires = [[1, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 1],
[0, 0, 1, 0, 0, 1, 0],
[0, 0, 0, 1, 1, 0, 0]]
pg = PygorRewire(pg, new_wires)
lb_short = lb_short[:4]
ub_short = ub_short[:4]
num_active_gates = len(new_wires)
'''
# choose the origin
if np.isscalar(origin):
origin = origin*np.ones(num_active_gates)
else:
if len(origin) != num_active_gates:
raise ValueError('Wrong array shape of the origin.')
origin.dump(str(save_dir / 'origin.npy'))
# important algorithm parameters
#threshold_low = 0.2 * (max_current - min_current) + min_current
step_back = 100 # important param for measuring d
len_after_pinchoff=50
num_samples = 100 # number of initial samples (by Latin hypercube design)
detector_pinchoff = util.PinchoffDetectorThreshold(threshold_low) # pichoff detector
detector_conducting = util.ConductingDetectorThreshold(threshold_high) # reverse direction
# create a Callable object, input: unit_vector, output: distance between the boundary and the origin
tester = Tester(pg, lb_short, ub_short, detector_pinchoff, d_r=d_r, len_after_pinchoff=len_after_pinchoff, logging=True, detector_conducting=detector_conducting, set_big_jump = set_big_jump, set_small_jump = set_small_jump)
###
# Set a problem and a model
###
#do_extra_meas = lambda vols, th: config_model.do_extra_meas(pg, vols, th)
do_extra_meas = None
###
# Gaussian process for r
###
l_prior_mean = 0.4 * np.ones(num_active_gates)
l_prior_var = 0.1*0.1 * np.ones(num_active_gates)
r_min, r_max = 0.0, np.sqrt(num_active_gates)* 2000.
v_prior_mean = ((r_max-r_min)/4.0)**2
#v_prior_var = v_prior_mean**2
#noise_var_r = np.square(d_r/2.0)
gp_r = GP_util.create_GP(num_active_gates, 'Matern52', v_prior_mean, l_prior_mean, (r_max-r_min)/2.0)
GP_util.set_GP_prior(gp_r, l_prior_mean, l_prior_var, None, None) # do not set prior for kernel var
GP_util.fix_hyperparams(gp_r, False, True)
###
# Gaussian process classifiers for predicting each probability component
###
l_prior_mean = 500. * np.ones(num_active_gates)
l_prior_var = 100.**2 * np.ones(num_active_gates)
v_prior_mean = 50.
v_prior_var = 20.**2
gpc_dict = dict()
gpc_dict['valid'] = GP_util.create_GP(num_active_gates, 'Matern52', lengthscale=l_prior_mean, const_kernel=True, GP=GPC)
gpc_dict['peak'] = GP_util.create_GP(num_active_gates, 'Matern52', lengthscale=l_prior_mean, const_kernel=True, GP=GPC) # when a point is valid
gpc_dict['goodscore'] = GP_util.create_GP(num_active_gates, 'Matern52', lengthscale=l_prior_mean, const_kernel=True, GP=GPC) # when a point is valid and has peaks
GP_util.set_GP_prior(gpc_dict['valid'], l_prior_mean, l_prior_var, v_prior_mean, v_prior_var) # do not set prior for kernel var
GP_util.set_GP_prior(gpc_dict['peak'], l_prior_mean, l_prior_var, v_prior_mean, v_prior_var) # do not set prior for kernel var
GP_util.set_GP_prior(gpc_dict['goodscore'], l_prior_mean, l_prior_var, v_prior_mean, v_prior_var) # do not set prior for kernel var
do_random_meas(num_active_gates, num_samples, tester, step_back, origin, lb_short, ub_short, gp_r, gpc_dict, do_extra_meas=do_extra_meas, save_dir=save_dir)
print('Time for random sampling: ', time.time()-tic)
delimiter=',',
fmt='%4.2f')
np.savetxt(save_dir / 'd_sorted.csv', d_all, delimiter=',', fmt='%4.2f')
np.savetxt(save_dir / 'dtot_sorted.csv',
d_tot[:, np.newaxis],
delimiter=',',
fmt='%4.2f')
#sys.exit()
# pygor
if conf_name != 'dummy':
pg, active_gate_idxs, lb, ub, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor(
conf_func, ip, settletime, settletime_big)
else:
pg, active_gate_idxs, lb, ub, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor_dummy(
shape)
active_gate_names = ["c{}".format(i + 1) for i in active_gate_idxs]
num_active_gates = len(active_gate_names)
print(active_gate_names)
# choose the origin
if np.isscalar(origin):
origin = origin * np.ones(num_active_gates)
else:
if len(origin) != num_active_gates:
raise ValueError('Wrong array shape of origin.')
# pinchoff detector
find_pinchoff_agian = False
if find_pinchoff_agian:
threshold = 0.2 * (max_current - min_current) + min_current
d_r = 2.5 # length of one step
def main():
conf_name = 'config1'
if conf_name == 'config1':
conf_func = config.config1
ip = "http://129.67.86.107:8000/RPC2"
save_dir = Path('./save_Dominic_redo_Basel2_correct')
settletime, settletime_big = 0.01, 0.03 # settletime_big is for shuttling, 'None' disables shuttling
# ['c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9', 'c10']
#origin = -100.
origin = 0.
threshold_low = 0.0 # for pinchoff detector
d_r = 10 # length of one step
elif conf_name == 'config2':
# ['c3', 'c4', 'c8', 'c10', 'c11', 'c12', 'c16']
conf_func = config.config2
ip = "http://129.67.85.235:8000/RPC2"
save_dir = Path('./save_Florian_redo')
settletime, settletime_big = 0.01, 0.03 # settletime_big is for shuttling, 'None' disables shuttling
#origin = -100.
origin = 0.
settletime = 0.01
threshold_low = 2.e-11
d_r = 10 # length of one step
elif conf_name == 'dummy':
import pygor_dummy
box_dim = 5
box_a = -1000. * np.ones(box_dim)
box_b = 1000. * np.ones(box_dim)
shape = pygor_dummy.Box(ndim=box_dim, a=box_a, b=box_b)
th_leak = np.array([-500., -400., -300., 0., 0.])
shape = pygor_dummy.Leakage(shape, th_leak)
save_dir = Path('./save_dummy')
origin = -100.
threshold_low = 0.2
d_r = 10 # length of one step
else:
raise ValueError('Unsupported setup')
save_dir.mkdir(exist_ok=True)
if conf_name != 'dummy':
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor(conf_func, ip, settletime, settletime_big)
else:
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor_dummy(shape)
threshold_high = 0.8 * max_current
active_gate_names = ["c{}".format(i+1) for i in active_gate_idxs]
num_active_gates = len(active_gate_names)
print(active_gate_names)
# choose the origin
if np.isscalar(origin):
origin = origin*np.ones(num_active_gates)
else:
if len(origin) != num_active_gates:
raise ValueError('Wrong array shape of origin.')
# important algorithm parameters
len_after_pinchoff=100
detector_pinchoff = util.PinchoffDetectorThreshold(threshold_low) # pichoff detector
detector_conducting = util.ConductingDetectorThreshold(threshold_high) # reverse direction
# create a Callable object, input: unit_vector, output: distance between the boundary and the origin
tester = Tester(pg, lb_short, ub_short, detector_pinchoff, d_r=d_r, len_after_pinchoff=len_after_pinchoff, logging=True, detector_conducting=detector_conducting, set_big_jump = set_big_jump, set_small_jump = set_small_jump)
#do_extra_meas = lambda vols: config_model.do_extra_meas(pg, vols)
do_extra_meas = None
# load voltages of a previous experiment
load_dir = Path('./save_Florian_randompoints')
#load_dir = Path('./save_Dominic_hs')
#load_dir = Path('./save_Dominic_randomHS_origin0_fixed')
vols_prev = np.load(load_dir/'vols_poff.npy', allow_pickle=True)
num_points = len(vols_prev)
vols_poff_all = list()
detected_all = list()
for i, v in enumerate(vols_prev):
v = convert_Basel2_to_Basel1(v)
v_origin = v - origin
u = v_origin / np.sqrt(np.sum(np.square(v_origin))) # voltage -> unit vector
# Get measurements
r, vols_pinchoff, found, t_firstjump = tester.get_r(u, origin=origin) # Measure the distance
vols_poff_all.append(vols_pinchoff)
detected_all.append(found)
print(i)
print(v)
print(vols_pinchoff)
np.array(vols_poff_all).dump(str(save_dir / 'vols_poff_after.npy'))
np.array(vols_prev).dump(str(save_dir / 'vols_poff_prev.npy'))
np.array(detected_all).dump(str(save_dir / 'detected_after.npy'))
def main():
conf_name = 'config2'
if conf_name == 'config1':
conf_func = config.config1
ip = "http://129.67.86.107:8000/RPC2"
save_dir = Path('./save_Dominic_ABL_group_optparam_run2/135_2')
settletime, settletime_big = 0.01, 0.03 # settletime_big is for shuttling, 'None' disables shuttling
# ['c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9', 'c10']
threshold_low = 0.0 # for pinchoff detector
d_r = 10 # length of one step
elif conf_name == 'config2':
# ['c3', 'c4', 'c8', 'c10', 'c11', 'c12', 'c16']
conf_func = config.config2
#ip = "http://129.67.85.235:8000/RPC2"
ip = 'http://129.67.85.38:8000/RPC2'
save_dir = Path('./save_Basel2_group/105')
settletime, settletime_big = 0.02, 0.02 # settletime_big is for shuttling, 'None' disables shuttling
threshold_low = 2.e-11
d_r = 10 # length of one step
elif conf_name == 'dummy':
import pygor_dummy
box_dim = 5
box_a = -1000. * np.ones(box_dim)
box_b = 1000. * np.ones(box_dim)
shape = pygor_dummy.Box(ndim=box_dim, a=box_a, b=box_b)
th_leak = np.array([-500., -400., -300., 0., 0.])
shape = pygor_dummy.Leakage(shape, th_leak)
save_dir = Path('./save_dummy')
threshold_low = 0.2
d_r = 10 # length of one step
else:
raise ValueError('Unsupported setup')
save_dir.mkdir(exist_ok=True)
if conf_name != 'dummy':
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor(
conf_func, ip, settletime, settletime_big)
else:
pg, active_gate_idxs, lb_short, ub_short, max_current, min_current, set_big_jump, set_small_jump = config.setup_pygor_dummy(
shape)
threshold_high = 0.8 * max_current
active_gate_names = ["c{}".format(i + 1) for i in active_gate_idxs]
num_active_gates = len(active_gate_names)
print(active_gate_names)
# important algorithm parameters
len_after_pinchoff = 100
detector_pinchoff = util.PinchoffDetectorThreshold(
threshold_low) # pichoff detector
detector_conducting = util.ConductingDetectorThreshold(
threshold_high) # reverse direction
# create a Callable object, input: unit_vector, output: distance between the boundary and the origin
tester = Tester(pg,
lb_short,
ub_short,
detector_pinchoff,
d_r=d_r,
len_after_pinchoff=len_after_pinchoff,
logging=True,
detector_conducting=detector_conducting,
set_big_jump=set_big_jump,
set_small_jump=set_small_jump)
#do_extra_meas = lambda vols, th: config_model.do_extra_meas(pg, vols, th)
do_extra_meas = None
#v_target = np.array([-1877.11448379, -315.09271412, -588.03713859, -1743.6191362 , -587.8001855 , -693.38667762, -832.91995001, -186.99101894])
#new_wires = [[1, 0, 0, 0, 0, 0, 0, 0],
#[0, 1, 0, 0, 0, 0, 0, 1],
#[0, 0, 1, 0, 0, 0, 1, 0],
#[0, 0, 0, 1, 1, 1, 0, 0]]
new_wires = [[1, 0, 0, 0, 0, 0, 0], [0, 1, 0, 0, 0, 0, 1],
[0, 0, 1, 0, 0, 1, 0], [0, 0, 0, 1, 1, 0, 0]]
new_wires = np.array(new_wires)
v_target = np.array(
[-1290.10731323, -1521.90819125, -813.40589252, -1069.5955569])
v_target = np.matmul(new_wires.T, v_target)
origin = np.minimum(v_target + 200., 0.)
print('Origin: ', origin)
np.array(v_target).dump(str(save_dir / 'v_local.npy'))
np.array(origin).dump(str(save_dir / 'origin_local.npy'))
num_trials = 500
sigma = 100.
#name_g1, name_g2 = 'c5', 'c9'
name_g1, name_g2 = 'c8', 'c12'
resol = 64
w1 = w2 = 75
idx_g1, idx_g2 = active_gate_names.index(name_g1), active_gate_names.index(
name_g2)
lb_g1, lb_g2 = lb_short[idx_g1], lb_short[idx_g2]
ub_g1, ub_g2 = ub_short[idx_g1], ub_short[idx_g2]
v_all = list()
meas_all = list()
time_all = list()
for i in range(num_trials):
'''
delta = np.random.normal(scale=sigma, size=num_active_gates)
v_origin = v_target + delta - origin
print('Random vol:', v_origin+origin)
u = v_origin / np.sqrt(np.sum(np.square(v_origin))) # voltage -> unit vector
'''
u = random_hypersphere(num_active_gates, 1)[0]
# Get measurements
t = time.time()
r, vols_pinchoff, found, t_firstjump = tester.get_r(
u, origin=origin) # Measure the distance
t1 = time.time() - t
print('Poff at ', vols_pinchoff)
if found:
# measurement
print('Found')
meas = measure_2d.scan2d(pg,
name_g1,
name_g2,
resol,
vols_pinchoff,
w1,
w2,
lb_g1,
lb_g2,
ub_g1,
ub_g2,
str(i),
pic_dpi=None,
mult1=1.,
mult2=1.,
save_dir=save_dir)
'''
meas = do_extra_meas(vols_pinchoff, 0.0)
# highres = meas[-6] if len(meas) > 10 else None
# lowres = meas[4] if len(meas) > 4 else None
if lowres:
plt.imsave(str(save_dir/(str(i)+'lowres.png')), lowres, origin='lower', cmap='viridis')
if highres:
plt.imsave(str(save_dir/(str(i)+'highres.png')), highres, origin='lower', cmap='viridis')
'''
else:
print('Not found')
meas = []
t2 = time.time() - t
time_all.append((t1, t2))
v_all.append(vols_pinchoff)
meas_all.append(meas)
np.array(v_all).dump(str(save_dir / 'vols_poff.npy'))
#np.array(meas_all).dump(str(save_dir/'meas.npy'))
np.save(str(save_dir / 'meas'), meas_all)
np.array(time_all).dump(str(save_dir / 'time.npy'))