def compute_xmax_cnet(cnet_module, depth, dataset, x_var, e_var):
x_map_list = [] # store the map tuple
evid_flag = np.zeros(dataset.shape[1],
dtype=int) # value = 1 means evidence
evid_flag[e_var] = 1
cnet = CNET_dfs(cnet_module, dataset.shape[1])
internal_list, leaf_list = cnet.get_node_list()
for t in leaf_list:
t.evar = np.where(evid_flag[t.ids] == 1)[0]
t.ind_as_parent(t.evar)
t.ind_as_child(t.evar)
xmax_prob_arr = np.zeros(dataset.shape[0])
for i in xrange(dataset.shape[0]):
datapoint = dataset[i]
xmax_prob_arr[i], x_map = cnet.instantiation(datapoint, e_var,
evid_flag, internal_list,
leaf_list)
x_map_list.append(x_map)
return xmax_prob_arr, np.asarray(x_map_list)
def compute_xmax_bcnet(bcnet, dataset, x_var, e_var):
x_map_list = [] # store the map tuple
evid_flag = np.zeros(dataset.shape[1],
dtype=int) # value = 1 means evidence
evid_flag[e_var] = 1
n_variables = dataset.shape[1]
cm_weights = bcnet['cm_weights']
log_weights = np.log(cm_weights)
n_components = cm_weights.shape[0]
bcnet_dict_list = []
bcnet_internal_list = []
bcnet_leaf_list = []
for i in xrange(n_components):
cn = bcnet['cnet_list'][i]
cnet_component = CNET_dfs(cn, n_variables)
bcnet_dict_list.append(cnet_component)
internal_list, leaf_list = cnet_component.get_node_list()
bcnet_internal_list.append(internal_list)
bcnet_leaf_list.append(leaf_list)
for i in xrange(len(bcnet_leaf_list)):
for t in bcnet_leaf_list[i]:
t.evar = np.where(evid_flag[t.ids] == 1)[0]
t.ind_as_parent(t.evar)
t.ind_as_child(t.evar)
xmax_prob_arr = np.zeros(dataset.shape[0])
for i in xrange(dataset.shape[0]):
datapoint = dataset[i]
xmap_temp = []
xmax_prob_temp = np.zeros(n_components)
for j in xrange(n_components):
cnet = bcnet_dict_list[j]
xmax_prob_temp[j], x_map = cnet.instantiation(
datapoint, e_var, evid_flag, bcnet_internal_list[j],
bcnet_leaf_list[j])
xmap_temp.append(x_map)
xmax_prob_temp += log_weights
best_ind = np.argmax(xmax_prob_temp)
# find the best one
x_map_list.append(xmap_temp[best_ind])
xmax_prob_arr[i] = xmax_prob_temp[best_ind]
return xmax_prob_arr, np.asarray(x_map_list)
def structure_redefine(self, load_info):
self.weights = load_info.mixture_weight
self.n_components = load_info.n_components
for cn in load_info.cnet_list:
main_dict = {}
utilM.save_cutset(main_dict,
cn.tree,
np.arange(self.n_variable),
ccpt_flag=True)
cnet_component = CNET_dfs(main_dict, self.n_variable)
self.cnet_dict_list.append(cnet_component)
internal_list, leaf_list = cnet_component.get_node_list()
self.internal_list.append(internal_list)
self.leaf_list.append(leaf_list)
def compute_xmax_bcnet(bcnet, dataset, x_var, e_var):
#print ('In compute_xmax_bcnet')
x_map_list = [] # store the map tuple
evid_flag = np.zeros(dataset.shape[1],
dtype=int) # value = 1 means evidence
evid_flag[e_var] = 1
n_variables = dataset.shape[1]
# cm_weights = np.loadtxt(bcnet_dir +'_component_weights.txt')
# log_weights = np.log(cm_weights)
#
# n_components = cm_weights.shape[0]
#
# bcnet_dict_list = []
# bcnet_internal_list =[]
# bcnet_leaf_list =[]
# for i in xrange(n_components):
# cn_file = bcnet_dir +'_' +str(i) + '.npz'
# cn = np.load(cn_file)['module'].item()
# cnet_component = CNET_dfs(cn,n_variables)
# bcnet_dict_list.append(cnet_component)
#
# internal_list, leaf_list = cnet_component.get_node_list()
# bcnet_internal_list.append(internal_list)
# bcnet_leaf_list.append(leaf_list)
cm_weights = bcnet['cm_weights']
log_weights = np.log(cm_weights)
n_components = cm_weights.shape[0]
bcnet_dict_list = []
bcnet_internal_list = []
bcnet_leaf_list = []
for i in xrange(n_components):
cn = bcnet['cnet_list'][i]
cnet_component = CNET_dfs(cn, n_variables)
bcnet_dict_list.append(cnet_component)
internal_list, leaf_list = cnet_component.get_node_list()
bcnet_internal_list.append(internal_list)
bcnet_leaf_list.append(leaf_list)
#print ('log_weights: ', log_weights)
for i in xrange(len(bcnet_leaf_list)):
for t in bcnet_leaf_list[i]:
t.evar = np.where(evid_flag[t.ids] == 1)[0]
t.ind_as_parent(t.evar)
t.ind_as_child(t.evar)
xmax_prob_arr = np.zeros(dataset.shape[0])
for i in xrange(dataset.shape[0]):
datapoint = dataset[i]
xmap_temp = []
xmax_prob_temp = np.zeros(n_components)
for j in xrange(n_components):
cnet = bcnet_dict_list[j]
xmax_prob_temp[j], x_map = cnet.instantiation(
datapoint, e_var, evid_flag, bcnet_internal_list[j],
bcnet_leaf_list[j])
xmap_temp.append(x_map)
#print ('xmax_prob_temp:', xmax_prob_temp)
xmax_prob_temp += log_weights
#print ('xmax_prob_temp:', xmax_prob_temp)
best_ind = np.argmax(xmax_prob_temp)
#print ('best_ind: ', best_ind)
# find the best one
x_map_list.append(xmap_temp[best_ind])
xmax_prob_arr[i] = xmax_prob_temp[best_ind]
return xmax_prob_arr, np.asarray(x_map_list)