def testGlobCoupPwBlrWithCpsParentMoves(data):
output_line = (
'Globally Coupled Bayesian Piece-Wise Linear Regression with moves on '
+ 'change-points and parent sets on mTOR Data.')
print(output_line)
logger.info(output_line) # Print and write output
baNet = Network(data, args.chain_length, args.burn_in, args.lag)
baNet.infer_network('glob_coup_nh_dbn')
# save the chain into the output folder
save_chain('mTOR_glob_coup_dbn.pckl', baNet)
return baNet.proposed_adj_matrix
def testVvGlobCoup(data, true_inc):
output_line = (
'Varying Variances Globally Coupled Bayesian Piece-Wise Linear Regression with moves on '
+ 'change-points and parent sets on Yeast Data.')
print(output_line)
logger.info(output_line) # Print and write output
baNet = Network(data, args.chain_length, args.burn_in)
baNet.infer_network('var_glob_coup_nh_dbn')
adjMatrixRoc(baNet.proposed_adj_matrix, true_inc, args.verbose)
# save the chain into the output folder
save_chain('vv_glob_coup_dbn.pckl', baNet)
def testSeqCoupPwBlrWithCpsParentMoves(data, true_inc):
output_line = (
'Sequentially Coupled Bayesian Piece-Wise Linear Regression with moves on '
+ 'change-points and parent sets on Yeast data.')
print(output_line)
logger.info(output_line) # Print and write output
baNet = Network(data, args.chain_length, args.burn_in)
baNet.infer_network('seq_coup_nh_dbn')
adjMatrixRoc(baNet.proposed_adj_matrix, true_inc, args.verbose)
# save the chain into the output folder
save_chain('seq_coup_dbn.pckl', baNet)
def main():
# The coefficients that will be used to generate the random data
coefs = parseCoefs(args.coefs_file)
# Select and run the chosen algorithm
if args.method == 'h-dbn':
func = test_h_dbn # Uncomment for testing the second algo on a network
elif args.method == 'nh-dbn':
func = testPwBlrWithCpsParentMoves # Test the fourth algorithm
elif args.method == 'seq-dbn':
func = testSeqCoupPwBlrWithCpsParentMoves # test the fifth algorithm
elif args.method == 'glob-dbn':
func = testGlobCoupPwBlrWithCpsParentMoves # test the sixth algorithm
elif args.method == 'var-glob-dbn':
func = testVvGlobCoup
with concurrent.futures.ThreadPoolExecutor() as executor:
# Generate data to test our algo
network, _, adjMatrix = generateNetwork(args.num_features,
args.num_indep, coefs,
args.num_samples,
args.change_points.copy(),
args.verbose,
args.generated_noise_var)
chain_number = [x + 1 for x in range(args.number_chains)]
networks_vec = [
copy.deepcopy(network) for _ in range(args.number_chains)
]
adjMatrix_vec = [
copy.deepcopy(adjMatrix) for _ in range(args.number_chains)
]
results = [
executor.submit(func, args)
for args in zip(chain_number, networks_vec, adjMatrix_vec)
]
for f in concurrent.futures.as_completed(results):
print(f.result()) # debug to check if the result was computed
baNet, file_name, flattened_true, flattened_scores = f.result(
) # deconstruct future output
adjMatrixRoc(
flattened_scores, flattened_true,
args.verbose) # cannot display or save on a non-main thread
# save the chain into the output folder
save_chain(file_name, baNet)
def testGlobCoupPwBlrWithCpsParentMoves(data, true_inc):
output_line = (
'Full Parents Credible Intervals Globally Coupled Bayesian Piece-Wise Linear Regression' +
'with moves on change-points only Yeast Data.'
)
print(output_line) ; logger.info(output_line) # Print and write output
baNet = Network(data, args.chain_length, args.burn_in, args.lag)
baNet.infer_network('fp_glob_coup_nh_dbn')
flattened_true, flattened_scores = transformResults(true_inc, baNet.proposed_adj_matrix)
adjMatrixRoc(flattened_scores, flattened_true, args.verbose)
# save the chain into the output folder
save_chain('fp_glob_coup_dbn.pckl', baNet)
def testPwBlrWithCpsParentMoves(data, true_inc):
output_line = ('Bayesian Piece-Wise Linear Regression with moves on ' +
'change-points and parent sets for the Yeast data.')
print(output_line)
logger.info(output_line) # Print and write output
if args.change_points == 0:
args.change_points = []
args.change_points.append(data.shape[0] +
1) # append the len data + 1 so the algo works
baNet = Network(data, args.chain_length, args.burn_in)
baNet.infer_network('varying_nh_dbn')
adjMatrixRoc(baNet.proposed_adj_matrix, true_inc, args.verbose)
# save the chain into the output folder
save_chain('nh_dbn.pckl', baNet)
def test_h_dbn(data, true_inc):
output_line = (
'Homogeneous Dinamic Bayesian Linear Regression with full parents ' +
'for the Yeast data. \n'
)
print(output_line) ; logger.info(output_line) # Print and write output
change_points = [] # set the cps empty list because this is the homegeneous version
# Create/Call the Network objects/methods
baNet = Network(data, args.chain_length, args.burn_in, args.lag, args.change_points) # Create theh BN obj
baNet.infer_network('fp_h_dbn') # Do the fixed parents version of the DBN algo
# trueAdjMatrix = adjMatrix[0] # For the moment we just get the adj matrix of the first cp
flattened_true, flattened_scores = transformResults(true_inc, baNet.proposed_adj_matrix)
adjMatrixRoc(flattened_scores, flattened_true, args.verbose)
# save the chain into the output folder
save_chain('fp_h_dbn.pckl', baNet)
def test_h_dbn(data):
output_line = ('Bayesian Linear Regression with moves on ' +
'the parent set only for the mTOR data. \n')
print(output_line)
logger.info(output_line) # Print and write output
change_points = [
] # set the cps empty list because this is the homegeneous version
# Create/Call the Network objects/methods
baNet = Network(data, args.chain_length, args.burn_in, args.lag,
args.change_points) # Create theh BN obj
baNet.infer_network(
'h_dbn') # Do the fixed parents version of the DBN algo
# save the chain into the output folder
save_chain('mTOR_h_dbn.pckl', baNet)
return baNet.proposed_adj_matrix
def test_h_dbn(data, true_inc):
output_line = ('Bayesian Linear Regression with moves on ' +
'the parent set only for the Yeast data. \n')
print(output_line)
logger.info(output_line) # Print and write output
change_points = [
] # set the cps empty list because this is the homegeneous version
# Create/Call the Network objects/methods
baNet = Network(data, args.chain_length, args.burn_in,
args.change_points) # Create theh BN obj
baNet.infer_network(
'h_dbn') # Do the fixed parents version of the DBN algo
# trueAdjMatrix = adjMatrix[0] # For the moment we just get the adj matrix of the first cp
adjMatrixRoc(baNet.proposed_adj_matrix, true_inc, args.verbose)
# save the chain into the output folder
save_chain('h_dbn.pckl', baNet)
def testPwBlrWithCpsParentMoves(data):
output_line = ('Bayesian Piece-Wise Linear Regression with moves on ' +
'change-points and parent sets for the mTOR data.')
print(output_line)
logger.info(output_line) # Print and write output
if args.change_points == 0:
args.change_points = []
# to get the total length of the data over the segments
# data_len = 0
# for segment in data:
# data_len = data_len + segment.shape[0]
# args.change_points.append(data_len + 1) # append the len data + 1 so the algo works
baNet = Network(data, args.chain_length, args.burn_in, args.lag)
baNet.infer_network('varying_nh_dbn')
# save the chain into the output folder
save_chain('mTOR_nh_dbn.pckl', baNet)
return baNet.proposed_adj_matrix
def testPwBlrWithCpsParentMoves(data, true_inc):
# credible intervals computation with full parents
output_line = (
'Bayesian Piece-Wise Linear Regression with moves on ' +
'change-points and full parent set for the Yeast data.'
)
print(output_line) ; logger.info(output_line) # Print and write output
if args.change_points == 0:
args.change_points = []
# to get the total length of the data over the segments
data_len = 0
for segment in data:
data_len = data_len + segment.shape[0]
args.change_points.append(data_len + 1) # append the len data + 1 so the algo works
baNet = Network(data, args.chain_length, args.burn_in, args.lag)
baNet.infer_network('fp_varying_nh_dbn')
flattened_true, flattened_scores = transformResults(true_inc, baNet.proposed_adj_matrix)
adjMatrixRoc(flattened_scores, flattened_true, args.verbose)
# save the chain into the output folder
save_chain('fp_nh_dbn.pckl', baNet)
