Python Optimizer 예제들

예제 #1

파일: test_optimize_diffusion_leftboundarydata.py 프로젝트: ebilionis/variational-reformulation-of-inverse-problems

# Pick an entropy approximation
entropy = FirstOrderEntropyApproximation()
# Pick an approximation for the expectation of the joint
expectation_functional = ThirdOrderExpectationFunctional(log_p)
# Restrictions for mu
mu_bounds = (tuple((0., 1.) for i in xrange(log_q.num_dim - 1))
            + ((1e-6, None), ))
C_bounds = tuple((1e-32, None) for i in xrange(log_q.num_comp * log_q.num_dim))
# Build the ELBO
elbo = EvidenceLowerBound(entropy, expectation_functional)
print 'ELBO:'
print str(elbo)


# Optimize the elbo
optimizer = Optimizer(elbo)

results_file = os.path.join('demos', 'diffusion_left_new_2_cali.pcl')
if os.path.exists(results_file):
    print 'I found:', results_file
    print 'I am skipping the experiment.'
    print 'Delete the file if you want to repeat it.'
    with open(results_file, 'rb') as fd:
        results = pickle.load(fd)
    L = results['L']
    log_q = results['log_q']
else:
    L = optimizer.optimize(log_q, tol=1e-3, max_it=10, mu_bounds=mu_bounds,
                           mu_constraints=None, C_bounds=C_bounds)
    result = {}
    result['L'] = L

예제 #2

파일: test_optimize_diffusion_upperleft.py 프로젝트: tsilifis/variational-reformulation-of-inverse-problems

# Pick an entropy approximation
entropy = FirstOrderEntropyApproximation()
# Pick an approximation for the expectation of the joint
expectation_functional = ThirdOrderExpectationFunctional(log_p)
# Restrictions for mu
mu_bounds = (tuple((0., 1.) for i in xrange(log_q.num_dim - 1))
            + ((1e-6, None), ))
C_bounds = tuple((1e-32, None) for i in xrange(log_q.num_comp * log_q.num_dim))
# Build the ELBO
elbo = EvidenceLowerBound(entropy, expectation_functional)
print 'ELBO:'
print str(elbo)


# Optimize the elbo
optimizer = Optimizer(elbo)

results_file = os.path.join('demos', 'diffusion_upleft_cali.pcl')
if os.path.exists(results_file):
    print 'I found:', results_file
    print 'I am skipping the experiment.'
    print 'Delete the file if you want to repeat it.'
    with open(results_file, 'rb') as fd:
        results = pickle.load(fd)
    L = results['L']
    log_q = results['log_q']
else:
    L = optimizer.optimize(log_q, tol=1e-3, max_it=10, mu_bounds=mu_bounds,
                           mu_constraints=None, C_bounds=C_bounds)
    result = {}
    result['L'] = L

예제 #3

파일: test_optimize_catalysis.py 프로젝트: tsilifis/variational-reformulation-of-inverse-problems

# Pick an entropy approximation
entropy = FirstOrderEntropyApproximation()
# Pick an approximation for the expectation of the joint
expectation_functional = ThirdOrderExpectationFunctional(log_p)
# Restrictions for mu
mu_bounds = (tuple((1e-6, 1) for i in xrange(log_q.num_dim - 1))
            + ((1e-6, None), ))
C_bounds = tuple((1e-32, None) for i in xrange(log_q.num_comp * log_q.num_dim))
# Build the ELBO
elbo = EvidenceLowerBound(entropy, expectation_functional)
print 'ELBO:'
print str(elbo)

# Optimize the elbo
optimizer = Optimizer(elbo)

results_file = os.path.join('demos', 'catalysis_1_cali.pcl')
if os.path.exists(results_file):
    print 'I found:', results_file
    print 'I am skipping the experiment.'
    print 'Delete the file if you want to repeat it.'
    with open(results_file, 'rb') as fd:
        results = pickle.load(fd)
    L = results['L']
    log_q = results['log_q']
else:
    L = optimizer.optimize(log_q, max_it=10, mu_bounds=mu_bounds,
                           C_bounds=C_bounds)
    result = {}
    result['L'] = L

예제 #4

파일: test_optimize_catalysis_prop_noise.py 프로젝트: ebilionis/variational-reformulation-of-inverse-problems

# Pick an entropy approximation
entropy = FirstOrderEntropyApproximation()
# Pick an approximation for the expectation of the joint
expectation_functional = ThirdOrderExpectationFunctional(log_p)
# Restrictions for mu
mu_bounds = (tuple((1e-6, None) for i in xrange(log_q.num_dim - 1))
            + ((1e-6, None), ))
C_bounds = tuple((1e-10, None) for i in xrange(log_q.num_comp * log_q.num_dim))
# Build the ELBO
elbo = EvidenceLowerBound(entropy, expectation_functional)
print 'ELBO:'
print str(elbo)

# Optimize the elbo
optimizer = Optimizer(elbo)

results_file = os.path.join('demos', 'catalysis_prop_noise_cali.pcl')
if os.path.exists(results_file):
    print 'I found:', results_file
    print 'I am skipping the experiment.'
    print 'Delete the file if you want to repeat it.'
    with open(results_file, 'rb') as fd:
        results = pickle.load(fd)
    L = results['L']
    log_q = results['log_q']
else:
    L = optimizer.optimize(log_q, max_it=10, mu_bounds=mu_bounds,
                           C_bounds=C_bounds)
    result = {}
    result['L'] = L

예제 #5

파일: test_optimize_catalysis_full_dmnl.py 프로젝트: ebilionis/variational-reformulation-of-inverse-problems

n = int(math.sqrt(log_q.num_dim - 1))
for j in xrange(n):
    c = {}
    c['type'] = 'eq'
    c['fun'] = mu_c_fun
    c['jac'] = mu_c_jac
    c['args'] = (j, )
    mu_constraints.append(c)

# Build the ELBO
elbo = EvidenceLowerBound(entropy, expectation_functional)
print 'ELBO:'
print str(elbo)

# Optimize the elbo
optimizer = Optimizer(elbo)

results_file = os.path.join('demos', 'catalysis_full_dmnl_cali.pcl')
if os.path.exists(results_file):
    print 'I found:', results_file
    print 'I am skipping the experiment.'
    print 'Delete the file if you want to repeat it.'
    with open(results_file, 'rb') as fd:
        results = pickle.load(fd)
    L = results['L']
    log_q = results['log_q']
else:
    L = optimizer.optimize(log_q, max_it=10, mu_bounds=mu_bounds,
                           mu_constraints=mu_constraints,
                           C_bounds=C_bounds)
    result = {}