def load_median_experiments(ini_file_path,date_exp):
Config.read(ini_file_path)
Config.sections()
models = dict(Config.items('inference'))
number_data_points = ConfigSectionMap("test-data")['data-points'].split(',')
list_noise_variance = ConfigSectionMap("test-data")['observation_noise'].split(',')
repeat = ConfigSectionMap("others")['repeat']
number_test_points = ConfigSectionMap("others")['number-test-points']
test_problems = ConfigSectionMap("test-data")['test-problems'].split(',')
total_steps_outer = ConfigSectionMap("MCMC")["total-steps-outer"]
directory="results/"+date_exp
residual_list = []
df_list= []
index_it=0
for key in models:
all_inf_string = ConfigSectionMap("inference")[key]
list_of_infer= all_inf_string.split(";")
for infer in list_of_infer:
for test_problem in test_problems:
for noise in list_noise_variance:
for n in number_data_points:
df = pd.DataFrame()
print(index_it)
index_it+=1
for index in range(int(repeat)):
experiment_name = key+'_'+infer+'_'+total_steps_outer+'_'\
+ test_problem +'_'+noise+'_'+n+'_'+number_test_points+'_'+str(index)
output_file_name = directory+"/exp_"+ experiment_name
try:
df = pd.read_pickle(output_file_name)
residual_list.append(np.abs(df['residuals'].values))
except ValueError:
("could not open "+output_file_name)
res_matrix = np.matrix(residual_list)
residual_list = []
df['median-residual']=pd.Series([np.median(res_matrix[:,i].tolist()) for i in df.index ])
df['model'] = pd.Series([key for _ in range(len(df.index))], index=df.index)
df['test_problem'] = pd.Series([test_problem for _ in range(len(df.index))], index=df.index)
df['noise'] = pd.Series([noise for _ in range(len(df.index))], index=df.index)
df['n'] = pd.Series([n for _ in range(len(df.index))], index=df.index)
df['repeat'] = pd.Series([index for _ in range(len(df.index))], index=df.index)
df_list.append(df)
df_experiment=pd.concat(df_list)
df_experiment.to_pickle("results/median_residual_"+date_exp)
return df_experiment
def load_experiments(ini_file_path,date_exp):
Config.read(ini_file_path)
Config.sections()
models = dict(Config.items('inference'))
number_data_points = ConfigSectionMap("test-data")['data-points'].split(',')
list_noise_variance = ConfigSectionMap("test-data")['observation_noise'].split(',')
repeat = ConfigSectionMap("others")['repeat']
number_test_points = ConfigSectionMap("others")['number-test-points']
test_problems = ConfigSectionMap("test-data")['test-problems'].split(',')
total_steps_outer = ConfigSectionMap("MCMC")["total-steps-outer"]
condition=[]
directory="results/"+date_exp
un_averaged_frames=[]
for index in range(int(repeat)):
frames =[]
for key in models:
all_inf_string = ConfigSectionMap("inference")[key]
list_of_infer= all_inf_string.split(";")
for infer in list_of_infer:
for test_problem in test_problems:
for noise in list_noise_variance:
for n in number_data_points:
experiment_name = key+'_'+infer+'_'+total_steps_outer+'_'\
+ test_problem +'_'+noise+'_'+n+'_'+number_test_points+'_'+str(index)
output_file_name = directory+"/exp_"+ experiment_name
try:
df = pd.read_pickle(output_file_name)
df['model'] = pd.Series([key for _ in range(len(df.index))], index=df.index)
df['test_problem'] = pd.Series([test_problem for _ in range(len(df.index))], index=df.index)
df['noise'] = pd.Series([noise for _ in range(len(df.index))], index=df.index)
df['n'] = pd.Series([n for _ in range(len(df.index))], index=df.index)
df['repeat'] = pd.Series([index for _ in range(len(df.index))], index=df.index)
#df['mcmcm-step-index'] = pd.Series([i for i in range(len(total_steps_outer))], index=df.index)
frames.append(df)
except ValueError:
("could not open "+output_file_name)
un_averaged_frames.append(pd.concat(frames))
df_experiment = average_frames(un_averaged_frames)
df_experiment.to_pickle("results/experiment_"+date_exp)
return df_experiment
help="Do not use parallelism")
parser.add_argument("--cores", type=int, default=60,
help="Number of cores")
parser.add_argument("-m", metavar="MESSAGE", type=str, default=" no message ", help="Message")
ns = parser.parse_args()
ini_file_path = ns.ini_file_path
date_exp = ns.d
run_locally = ns.local
cores = ns.cores
message = ns.m
Config.read(ini_file_path)
Config.sections()
models = dict(Config.items('inference'))
number_data_points = ConfigSectionMap("test-data")['data-points'].split(',')
list_noise_variance = ConfigSectionMap("test-data")['observation_noise'].split(',')
repeat = ConfigSectionMap("others")['repeat']
number_test_points = ConfigSectionMap("others")['number-test-points']
test_problems = ConfigSectionMap("test-data")['test-problems'].split(',')
total_steps_outer = ConfigSectionMap("MCMC")["total-steps-outer"]
condition=[]
for key in models:
all_inf_string = ConfigSectionMap("inference")[key]
list_of_infer= all_inf_string.split(";")
for infer in list_of_infer:
for test_problem in test_problems:
def get_last_n_residuals(ini_file_path,date_exp,n_last_residuals):
Config.read(ini_file_path)
Config.sections()
models = dict(Config.items('inference'))
number_data_points = ConfigSectionMap("test-data")['data-points'].split(',')
list_noise_variance = ConfigSectionMap("test-data")['observation_noise'].split(',')
repeat = ConfigSectionMap("others")['repeat']
number_test_points = ConfigSectionMap("others")['number-test-points']
test_problems = ConfigSectionMap("test-data")['test-problems'].split(',')
total_steps_outer = ConfigSectionMap("MCMC")["total-steps-outer"]
condition=[]
directory="results/"+date_exp
un_averaged_frames=[]
base_line = []
residuals=[]
model=[]
problem=[]
noise_level=[]
n_training_data=[]
for index in range(int(repeat)):
for key in models:
all_inf_string = ConfigSectionMap("inference")[key]
list_of_infer= all_inf_string.split(";")
for infer in list_of_infer:
for test_problem in test_problems:
for noise in list_noise_variance:
for n in number_data_points:
experiment_name = key+'_'+infer+'_'+total_steps_outer+'_'\
+ test_problem +'_'+noise+'_'+n+'_'+number_test_points+'_'+str(index)
output_file_name = directory+"/exp_"+ experiment_name
try:
df_all = pd.read_pickle(output_file_name)
df=df_all.tail(n_last_residuals)
base_line.extend(df['base-line'].iloc[0])
#index=[]
for j in range(len(df.index)):
residuals.extend(df['residuals'].iloc[j])
model.extend([key for _ in range(int(number_test_points))])
problem.extend([test_problem for _ in range(int(number_test_points))])
noise_level.extend([noise for _ in range(int(number_test_points))])
n_training_data.extend([n for _ in range(int(number_test_points))])
#index.append([j for _ in range(int(number_test_points))])
#residuals= np.concatenate(residuals,df['residuals'].iloc[j])
'''
df['model'] = pd.Series([key for _ in range(len(df.index))], index=df.index)
df['test_problem'] = pd.Series([test_problem for _ in range(len(df.index))], index=df.index)
df['noise'] = pd.Series([noise for _ in range(len(df.index))], index=df.index)
df['n'] = pd.Series([n for _ in range(len(df.index))], index=df.index)
df['repeat'] = pd.Series([index for _ in range(len(df.index))], index=df.index)
#df['mcmcm-step-index'] = pd.Series([i for i in range(len(total_steps_outer))], index=df.index)
'''
except ValueError:
("could not open "+output_file_name)
return pd.DataFrame({'residuals':residuals,'model':model,'test_problem':problem,'noise':noise_level,'n':n_training_data}),np.mean(base_line)
