def local_2(args):
"""Computes the SSE_local, SST_local and varX_matrix_local
Args:
args (dictionary): {"input": {
"avg_beta_vector": ,
"mean_y_global": ,
"computation_phase":
},
"cache": {
"covariates": ,
"dependents": ,
"lambda": ,
"dof_local": ,
}
}
Returns:
computation_output (json): {"output": {
"SSE_local": ,
"SST_local": ,
"varX_matrix_local": ,
"computation_phase":
}
}
Comments:
After receiving the mean_y_global, calculate the SSE_local,
SST_local and varX_matrix_local
"""
cache_list = args["cache"]
input_list = args["input"]
X = cache_list["covariates"]
y = cache_list["dependents"]
biased_X = sm.add_constant(X)
avg_beta_vector = input_list["avg_beta_vector"]
mean_y_global = input_list["mean_y_global"]
SSE_local = reg.sum_squared_error(biased_X, y, avg_beta_vector)
SST_local = np.sum(np.square(np.subtract(y, mean_y_global)))
varX_matrix_local = np.dot(biased_X.T, biased_X)
computation_output = {
"output": {
"SSE_local": SSE_local,
"SST_local": SST_local,
"varX_matrix_local": varX_matrix_local.tolist(),
"computation_phase": 'local_2'
},
"cache": {}
}
return json.dumps(computation_output)
def local_2(args, computation_phase):
"""Calculates the SSE_local, SST_local and varX_matrix_local
Args:
args (dictionary): {
'cache': {
'avg_beta_vector': ,
'mean_y_global': ,
'dof_global': ,
'dof_local': ,
'beta_vector_local': ,
'r_2_local': ,
'ts_local': ,
'ps_local':
},
'output': {
'avg_beta_vector': ,
'mean_y_global': ,
'computation_phase':
}
}
computation_phase (string): A field specifying which part
(local/remote) of the decentralized
computation has been performed last
In this case, it has to be remote_1
Returns:
computation_output (json): {
'output': {
'SSE_local': ,
'SST_local': ,
'varX_matrix_local': ,
'computation_phase':
}
}
Comments:
After receiving the mean_y_global, calculate the SSE_local,
SST_local and varX_matrix_local
"""
cache_list = args['cache']
input_list = args['input']
X = cache_list['covariates']
y = cache_list['dependents']
biased_X = np.insert(X, 0, 1, axis=1)
avg_beta_vector = input_list['avg_beta_vector']
mean_y_global = input_list['mean_y_global']
SSE_local = reg.sum_squared_error(biased_X, y, avg_beta_vector)
SST_local = np.sum(np.square(y - mean_y_global))
varX_matrix_local = np.dot(biased_X.T, biased_X)
computation_output_dict = {
'output': {
'SSE_local': SSE_local,
'SST_local': SST_local,
'varX_matrix_local': varX_matrix_local,
'computation_phase': computation_phase
}
}
return json.dumps(computation_output_dict,
sort_keys=True,
indent=4,
separators=(',', ': '))
def local_4(args):
"""Computes the SSE_local, SST_local and varX_matrix_local
Args:
args (dictionary): {"input": {
"avg_beta_vector": ,
"mean_y_global": ,
"computation_phase":
},
"cache": {
"covariates": ,
"dependents": ,
"lambda": ,
"dof_local": ,
}
}
Returns:
computation_output (json): {"output": {
"SSE_local": ,
"SST_local": ,
"varX_matrix_local": ,
"computation_phase":
}
}
Comments:
After receiving the mean_y_global, calculate the SSE_local,
SST_local and varX_matrix_local
"""
cache_list = args["cache"]
input_list = args["input"]
X = cache_list["covariates"]
y = cache_list["dependents"]
biased_X = np.array(X)
avg_beta_vector = input_list["avg_beta_vector"]
mean_y_global = input_list["mean_y_global"]
y = pd.DataFrame(y)
SSE_local, SST_local = [], []
for index, column in enumerate(y.columns):
curr_y = y[column].values
SSE_local.append(
reg.sum_squared_error(biased_X, curr_y, avg_beta_vector))
SST_local.append(
np.sum(
np.square(np.subtract(curr_y, mean_y_global[index])),
dtype=float))
varX_matrix_local = np.dot(biased_X.T, biased_X)
computation_output = {
"output": {
"SSE_local": SSE_local,
"SST_local": SST_local,
"varX_matrix_local": varX_matrix_local.tolist(),
"computation_phase": "local_4"
},
"cache": {}
}
return json.dumps(computation_output)
def local_2(args):
"""Computes the SSE_local, SST_local and varX_matrix_local
Args:
args (dictionary): {"input": {
"avg_beta_vector": ,
"mean_y_global": ,
"computation_phase":
},
"cache": {
"covariates": ,
"dependents": ,
"lambda": ,
"dof_local": ,
}
}
Returns:
computation_output (json): {"output": {
"SSE_local": ,
"SST_local": ,
"varX_matrix_local": ,
"computation_phase":
}
}
Comments:
After receiving the mean_y_global, calculate the SSE_local,
SST_local and varX_matrix_local
"""
cache_list = args["cache"]
input_list = args["input"]
X = pd.read_json(cache_list["covariates"], orient='split')
y = pd.read_json(cache_list["dependents"], orient='split')
biased_X = sm.add_constant(X.values)
avg_beta_vector = input_list["avg_beta_vector"]
mean_y_global = input_list["mean_y_global"]
SSE_local, SST_local, varX_matrix_local = [], [], []
for index, column in enumerate(y.columns):
curr_y = y[column]
X_, y_ = ignore_nans(biased_X, curr_y)
SSE_local.append(reg.sum_squared_error(X_, y_, avg_beta_vector[index]))
SST_local.append(
np.sum(np.square(np.subtract(y_, mean_y_global[index]))))
varX_matrix_local.append(np.dot(X_.T, X_).tolist())
output_dict = {
"SSE_local": SSE_local,
"SST_local": SST_local,
"varX_matrix_local": varX_matrix_local,
"computation_phase": 'local_2'
}
cache_dict = {}
computation_output = {"output": output_dict, "cache": cache_dict}
return json.dumps(computation_output)