def accuracy_VS_gamma(epsilon, prior, data, gammas):
mean_error = [[]]
for g in gammas:
Bayesian_Model = BayesInferwithDirPrior(prior, sum(data), epsilon, 0.1,
g)
Bayesian_Model._set_observation(data)
print("start" + str(g))
Bayesian_Model._experiments(1000)
print("finished" + str(g))
mean_error[0].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[3]])
print('Accuracy / prior: ' + str(prior._alphas) + ", delta: " +
str(delta) + ", epsilon:" + str(epsilon))
# print mean_error
plot_mean_error(gammas, mean_error, gammas,
"Different Gammas for Smooth Sensitivity",
[r"$\mathsf{EHDS}$"], "")
# plot_error_box(data,"Different Datasizes",datasizes,"Accuracy VS. Data Size",
# [r'$\mathcal{M}^{B}_{\mathcal{H}}$',"LapMech (sensitivity = 2)", "LapMech (sensitivity = 3)"],
# ['lightblue', 'navy', 'red'])
return
def accuracy_VS_prior(sample_size, epsilon, delta, priors, observation):
data = []
mean_error = [[], [], [], [], []]
for prior in priors:
Bayesian_Model = BayesInferwithDirPrior(prior, sample_size, epsilon,
delta)
Bayesian_Model._set_observation(observation)
Bayesian_Model._experiments(1000)
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[4]])
mean_error[0].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[3]])
mean_error[1].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[0]])
mean_error[2].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[4]])
print('Accuracy / observation: ' + str(observation) + ", delta: " +
str(delta) + ", epsilon:" + str(epsilon))
plot_error_box(data, r"Different Priors on $\theta$",
[r"$\mathsf{beta}$" + str(i._alphas) for i in priors],
"Accuracy VS. Prior Distribution", [
r'$\mathcal{M}_{\mathcal{H}}$',
"LapMech (sensitivity = 1)", "LapMech (sensitivity = 2)"
], ['navy', 'red', 'green'])
return
def run_experiments(times, datasizes, observations, epsilon, delta, prior):
data = []
errors = [[], [], [], [], []]
for i in range(len(datasizes)):
observation = observations[i]
Bayesian_Model = BayesInferwithDirPrior(prior, sum(observation),
epsilon, delta)
Bayesian_Model._set_observation(observation)
print("start" + str(observation))
Bayesian_Model._experiments(times)
print("finished" + str(observation))
for i in range(5):
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[i]])
plot_error_box(data, "Different Data Sets",
["bike", "cryotherapy", "immunotherapy"],
"Experiments on Real Data", [
r'Alg 1 - $\mathsf{LSDim}$ (sensitivity = 2.0)',
r'Alg 2 - $\mathsf{LSHist}$ (sensitivity = 1.0)',
r'Alg 5 - $\mathsf{EHDS}$ ', r"Alg 3 - $\mathsf{EHD}$",
r"Alg 4 - $\mathsf{EHDL}$"
], ["skyblue", "navy", "coral", "crimson", "blueviolet"])
return
def accuracy_VS_prior_mean(sample_size, epsilon, delta, priors, observations):
data = []
xlabel = []
for prior in priors:
for observation in observations:
Bayesian_Model = BayesInferwithDirPrior(prior, sample_size,
epsilon, delta)
Bayesian_Model._set_observation(observation)
Bayesian_Model._experiments(300)
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
xstick.append(
str(prior._alphas) + ", data:" + str(observation) + "/ExpMech")
xstick.append(
str(prior._alphas) + ", data:" + str(observation) + "/Laplace")
plot_error_box(data, "Different Prior Distributions", xstick,
"Accuracy VS. Prior Distribution")
return
def accuracy_VS_dimension(sample_sizes, epsilon, delta):
data = []
xstick = []
for n in sample_sizes:
for d in range(2, 5):
observation = [n for i in range(d)]
prior = Dir([1 for i in range(d)])
Bayesian_Model = BayesInferwithDirPrior(prior, n * d, epsilon,
delta)
Bayesian_Model._set_observation(observation)
Bayesian_Model._experiments(500)
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
xstick.append(str(observation) + "/ExpMech")
xstick.append(str(observation) + "/Laplace")
plot_error_box(data, "Different Prior Distributions", xstick,
"Accuracy VS. Prior Distribution")
return
def accuracy_VS_datasize(epsilons, delta, prior, observation, datasize):
data = []
mean_error = [[], []]
for e in epsilons:
Bayesian_Model = BayesInferwithDirPrior(prior, sum(observation), e,
delta, 0.2)
Bayesian_Model._set_observation(observation)
print("start" + str(observation))
Bayesian_Model._experiments(1000)
print("finished" + str(observation))
for j in range(len(mean_error)):
mean_error[j].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[j]])
plot_mean_error(epsilons, mean_error, [round(e, 2) for e in epsilons],
"Different Datasizes",
[r"$Laplace Noise$", r"$Geomoetric Noise$"], "")
return
def accuracy_VS_mean(sample_size, epsilon, delta, prior):
data = []
xstick = []
temp = BayesInferwithDirPrior(prior, sample_size, epsilon, delta)
temp._set_candidate_scores()
observations = temp._candidates
for i in range(len(observations)):
observations[i]._minus(prior)
for observation in observations:
Bayesian_Model = BayesInferwithDirPrior(prior, sample_size, epsilon,
delta)
Bayesian_Model._set_observation(observation._alphas)
Bayesian_Model._experiments(500)
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
xstick.append(str(observation._alphas) + "/ExpMech")
xstick.append(str(observation._alphas) + "/Laplace")
plot_error_box(data, "Different Prior Distributions", xstick,
"Accuracy VS. Prior Distribution")
return
def accuracy_VS_datasize(epsilon, delta, prior, observations, datasizes):
data = []
mean_error = [[], []]
for i in range(len(datasizes)):
observation = observations[i]
Bayesian_Model = BayesInferwithDirPrior(prior, sum(observation),
epsilon, delta, 0.2)
Bayesian_Model._set_observation(observation)
print("start" + str(observation))
Bayesian_Model._experiments(1000)
print("finished" + str(observation))
for j in range(len(mean_error)):
mean_error[j].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[j]])
print('Accuracy / prior: ' + str(prior._alphas) + ", delta: " +
str(delta) + ", epsilon:" + str(epsilon))
plot_mean_error(datasizes, mean_error, datasizes, "Different Datasizes",
[r"$Laplace Noise$", r"$Geomoetric Noise$"], "")
return
def accuracy_VS_datasize(epsilon, delta, prior, observations, datasizes):
data = []
mean_error = [[], [], [], [], [], []]
for i in range(len(datasizes)):
observation = observations[i]
Bayesian_Model = BayesInferwithDirPrior(prior, sum(observation),
epsilon, delta, 0.2)
Bayesian_Model._set_observation(observation)
print("start" + str(observation))
Bayesian_Model._experiments(500)
print("finished" + str(observation))
for j in range(len(mean_error)):
mean_error[j].append(
Bayesian_Model._accuracy_mean[Bayesian_Model._keys[j]])
# data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
# data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
# data.append(Bayesian_Model._accuracy[Bayesian_Model._keys[4]])
# a = statistics.median(Bayesian_Model._accuracy[Bayesian_Model._keys[3]])
# b = statistics.median(Bayesian_Model._accuracy[Bayesian_Model._keys[0]])
# c = statistics.median(Bayesian_Model._accuracy[Bayesian_Model._keys[4]])
print('Accuracy / prior: ' + str(prior._alphas) + ", delta: " +
str(delta) + ", epsilon:" + str(epsilon))
# print mean_error
plot_mean_error(datasizes, mean_error, datasizes, "Different Datasizes", [
r"$\mathsf{LSDim}$", r"$\mathsf{LSHist}$", r"$\mathsf{LSZhang}$",
r"$\mathsf{EHDS}$", r"$\mathsf{EHD}$", r"$\mathsf{EHDL}$"
], "")
# plot_error_box(data,"Different Datasizes",datasizes,"Accuracy VS. Data Size",
# [r'$\mathcal{M}^{B}_{\mathcal{H}}$',"LapMech (sensitivity = 2)", "LapMech (sensitivity = 3)"],
# ['lightblue', 'navy', 'red'])
return