def testVvGlobCoup(f_args):
thread_id, network, adjMatrix = f_args # deconstruct the f_arguments
start = time.perf_counter() # for timing the algorithm
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([network], args.chain_length, args.burn_in, args.lag)
baNet.infer_network('var_glob_coup_nh_dbn')
true_inc = 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)
finish = time.perf_counter()
print('My thread id is: ', thread_id, ' and I took: ',
round(finish - start, 2), ' to run.')
file_name = 'sim_seg_glob_coup_dbn_' + str(
thread_id) + '.pckl' # set filename that is based on thread id
return baNet, file_name, flattened_true, flattened_scores
def test_h_dbn(f_args):
thread_id, network, adjMatrix = f_args # deconstruct the f_arguments
start = time.perf_counter() # for timing the algorithm
output_line = ('Bayesian Linear Regression with moves on' +
'the parent set only. \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
baNet = Network([network], args.chain_length, args.burn_in, args.lag,
change_points) # Create theh BN obj
baNet.infer_network(
'h_dbn') # Do the fixed parents version of the DBN algo
true_inc = 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)
finish = time.perf_counter()
print('My thread id is: ', thread_id, ' and I took: ',
round(finish - start, 2), ' to run.')
file_name = 'sim_h_dbn_' + str(
thread_id) + '.pckl' # set filename that is based on thread id
return baNet, file_name, flattened_true, flattened_scores
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 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, args.lag)
baNet.infer_network('var_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('vv_glob_coup_dbn.pckl', baNet)
def main():
# true incidence for yeast data
# true_inc = [
# [0, 0, 1, 0, 1],
# [1, 0, 0, 1, 0],
# [0, 1, 0, 0 ,0],
# [0, 1, 1, 0, 0],
# [0, 0, 1, 0, 0]
# ]
# true incidence for simulated data
true_inc = [[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[1, 0, 1, 0, 0, 0], [1, 1, 0, 0, 0, 0], [1, 1, 1, 0, 0, 0]]
nmbr_chains = 5
model_name = 'sim_glob_coup_dbn'
chains = []
for idx in range(nmbr_chains):
file_name = model_name + '_' + str(idx + 1) + '.pckl'
curr_chain = load_chain(file_name)
_, flattened_scores = transformResults(true_inc,
curr_chain.proposed_adj_matrix)
chains.append(flattened_scores)
colors = ['tab:blue', 'tab:orange', 'tab:green', 'tab:red', 'tab:blue']
labels = ['chain_1', 'chain_2', 'chain_3', 'chain_4', 'chain_5']
fig, ax = plt.subplots()
for idx, edge_scores in enumerate(chains):
_x = [x + 1 for x in range(len(edge_scores))]
color = colors[idx]
label = labels[idx]
ax.scatter(_x,
edge_scores,
c=color,
label=label,
alpha=0.3,
edgecolors='none')
ax.xaxis.set_major_locator(
MaxNLocator(integer=True)) # force x ticks as integers
ax.legend()
ax.grid(True)
plt.xlabel('Edge Number')
plt.ylabel('Edge Score')
plt.show()
def testPwBlrWithParentMoves(data, true_inc):
output_line = (
'Bayesian Piece-Wise Linear Regression with moves on ' +
'the parent set only with fixed changepoints for the Yeast Data. \n')
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, args.lag,
args.change_points) # Create theh BN obj
baNet.infer_network(
'fixed_nh_dbn') # Do the fixed changepoints version of the DBN algo
flattened_true, flattened_scores = transformResults(
true_inc, baNet.proposed_adj_matrix)
adjMatrixRoc(flattened_scores, flattened_true, args.verbose)
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 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 = []
# 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')
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('nh_dbn.pckl', baNet)
