def start(logger):
setup = set_up()
model_param = get_model_param()
parameters = {'alpha': 0, 'beta': 1, 'g_dsa': 2, 'g_dsr': 3}
input_parameters = {'alpha': np.linspace(1.0, 2.5, 101), 'beta': np.linspace(0.02, 0.2, 101),
'g_dsa': np.linspace(0., 1., 101), 'g_dsr': np.linspace(0., 1., 101)}
for rand in range(30):
for biased_parameter, biased_idx in parameters.items():
result = dict()
logger.debug("Simulation: %s" % setup['expected'])
mu = setup['expected'][biased_idx]
input_ = input_parameters[biased_parameter]
irr_symbol = IrradianceModel(model_param['x'], model_param['zenith'], model_param['AMass'], model_param['pressure'], model_param['ssa'])
getIrrRatio = irr_symbol.getcompiledModel('ratio')
simulation = getIrrRatio(*setup['expected']) + np.random.normal(0, 0.001, len(model_param['x']))
expected = copy.copy(setup['expected'])
variables = copy.copy(setup['variables'])
guess = copy.copy(setup['guess'])
bounds = copy.copy(setup['bounds'])
del variables[biased_idx]
del expected[biased_idx]
del guess[biased_idx]
del bounds[biased_idx]
result['output'] = _iterate(simulation, variables, expected, guess, bounds, biased_parameter, irr_symbol, input_, logger)
result['input'] = input_ - mu
logger.info("======= Got %s" % result['output'])
result['expected'] = np.zeros(len(input_))
result['expected'][0:len(setup['expected'])] = setup['expected']
frame = pd.DataFrame(result['output'], columns=variables)
frame.insert(0, '%s' % biased_parameter, result['input'])
frame.insert(0, 'expected', result['expected'])
frame.to_csv('results/biased_%s_%s.txt' % (rand, biased_parameter), index=False)
def start2D(logger):
model_param = get_model_param()
setup = set_up()
r_setup = copy.copy(set_up())
noise_count = 40
global_var = {'param': 'g_dsr', 'input': np.linspace(.1,.5,101), 'idx':3}
local_var = {'param': 'g_dsa', 'input': np.linspace(.1,.5,101) ,'idx':2}
for key in setup.keys():
del r_setup[key][global_var['idx']]
del r_setup[key][local_var['idx']]
irr_symbol = IrradianceModel(model_param['x'], model_param['zenith'], model_param['AMass'], model_param['pressure'], model_param['ssa'])
getIrrRatio = irr_symbol.getcompiledModel('ratio')
simulation = getIrrRatio(*setup['expected'])
fit_parameters = dict((key,0) for key in r_setup['variables'])
map_parameters = {'alpha': 0, 'beta': 1}
for key, value in fit_parameters.items():
fit_parameters[key] = ['%s_%s' %(key, i) for i in range(noise_count)]
for i, input_value in enumerate(global_var['input']):
irr_symbol.setVariable(global_var['param'], input_value)
logger.info(">>>> Global Input for %s: %s" %(global_var['param'], input_value))
result = dict()
frame = pd.DataFrame()
for count in range(noise_count):
noised_simulation = simulation + np.random.normal(0, 0.001, len(model_param['x']))
logger.debug("Symbols %s" % irr_symbol.get_Symbols())
result['output'] = _iterate(noised_simulation, [], r_setup['expected'], r_setup['guess'], r_setup['bounds'], local_var['param'], irr_symbol, local_var['input'], logger, 2)
for key, value in fit_parameters.items():
logger.debug("Inlude for parameter %s the column %s at index %s" % (key, value[count], map_parameters[key]))
frame.insert(0, value[count], result['output'][:, map_parameters[key]])
for key, value in fit_parameters.items():
mean = frame[value].iloc[:].mean(axis=1)
std = frame[value].iloc[:].std(axis=1)
frame.insert(0, '%s_std' % key, std)
frame.insert(0, '%s_mean' % key, mean)
frame.insert(0, local_var['param'], local_var['input'] - setup['expected'][local_var['idx']])
buf = np.zeros(len(local_var['input']))
buf[0] = input_value - setup['expected'][global_var['idx']]
frame.insert(0, global_var['param'], buf)
buf[0:len(setup['expected'])] = setup['expected']
frame.insert(0, 'expected', buf)
frame.to_csv('results/two_variation_Deltaoutput_%s.txt' % i, index=False)
logger.info(frame)
logger.debug(r_setup)