def __init__(self, frecords, classifier, label_mappings, record_mappings,
get_labels_of_record):
'''
Constructor.
frecords - generator returning records.
NOTE: if a user wants to manipulate, which codes to consider(e.g. higher or lower level)
it is good to give a specific frecords parameter
classifier - constructor of a multi-label classifier, parametrized by frecords only.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
record_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map record into sub-record (in terms of labels) that is important in the current node-level; mapping should be specified with child_id for
additional filtering
IMPORTANT: SHOULD NOT MODIFY RECORD INTERNALLY, BUT RETURN A NEW RECORD!
'''
#derivable fields:
self.max_depth = len(label_mappings) - 1
self.get_labels_of_record = get_labels_of_record
#Build the classification tree.
PRINTER("[MlHierarchical: init]: start of training...")
self.mltree = tree_dfs_builder(
frecords, lambda records, curr_depth, child_id: self.node_builder(
records, curr_depth, child_id, classifier, record_mappings),
lambda records, curr_depth, child_id: self.children_splitter(
records, curr_depth, child_id, label_mappings), self.max_depth,
0, '') #we start at depth 0
def __init__(self, records, train_labels, classifier, label_mappings,
continue_deepening, is_leaf_node):
'''
Constructor.
records -records list.
train_labels - consecutive labels, for each record
classifier - constructor of a multi-label classifier, parametrized by records, train_labels and list_of_all_labels.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
continue_deepening - returns a boolean value, wether a given node has at least one non-leaf chid
is_leaf_node - returns a boolean value indicating wether a node is a leaf
'''
#derivable fields:
self.max_depth = len(label_mappings) - 1
#Build the classification tree.
PRINTER("[MlHierarchical: init]: start of training...")
self.continue_deepening = continue_deepening #wether to continue deepening from a child_id
self.is_leaf_node = is_leaf_node
self.mltree = tree_dfs_builder(
records, lambda records, curr_depth, child_id: self.node_builder(
records, curr_depth, child_id, classifier, label_mappings,
train_labels),
lambda records, curr_depth, child_id: self.children_splitter(
records, curr_depth, child_id, label_mappings, train_labels),
self.max_depth, 0, 'START_NODE',
self.continue_deepening) #we start at depth 0
def __init__(self, records, train_labels, classifier, label_mappings, continue_deepening, is_leaf_node):
'''
Constructor.
records -records list.
train_labels - consecutive labels, for each record
classifier - constructor of a multi-label classifier, parametrized by records, train_labels and list_of_all_labels.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
continue_deepening - returns a boolean value, wether a given node has at least one non-leaf chid
is_leaf_node - returns a boolean value indicating wether a node is a leaf
'''
#derivable fields:
self.max_depth = len(label_mappings)-1
#Build the classification tree.
PRINTER("[MlHierarchical: init]: start of training...")
self.continue_deepening = continue_deepening#wether to continue deepening from a child_id
self.is_leaf_node = is_leaf_node
self.mltree = tree_dfs_builder(records,
lambda records, curr_depth, child_id: self.node_builder(records, curr_depth, child_id, classifier, label_mappings, train_labels),
lambda records, curr_depth, child_id: self.children_splitter(records, curr_depth, child_id, label_mappings, train_labels), self.max_depth, 0, 'START_NODE', self.continue_deepening)#we start at depth 0
def __init__(self, frecords, classifier, label_mappings,
get_labels_of_record):
'''
Constructor.
frecords - generator returning records.
NOTE: if a user wants to manipulate, which codes to consider(e.g. higher or lower level)
it is good to give a specific frecords parameter
classifier - constructor of a multi-label classifier, parametrized by frecords and get_labels_of_record only.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
get_labels_of_record - returns a list of labels of the record
'''
#derivable fields:
self.max_depth = len(label_mappings) - 1
self.get_labels_of_record = get_labels_of_record
#Build the classification tree.
PRINTER("[init]: start of training...")
self.mltree = tree_dfs_builder(
frecords, lambda records, curr_depth, child_id: self.node_builder(
records, curr_depth, child_id, classifier, label_mappings,
get_labels_of_record),
lambda records, curr_depth, child_id: self.children_splitter(
records, curr_depth, child_id, label_mappings,
get_labels_of_record), self.max_depth, 0, '',
lambda x: True) #we start at depth 0
def __init__(self, frecords, classifier, label_mappings, record_mappings, get_labels_of_record):
'''
Constructor.
frecords - generator returning records.
NOTE: if a user wants to manipulate, which codes to consider(e.g. higher or lower level)
it is good to give a specific frecords parameter
classifier - constructor of a multi-label classifier, parametrized by frecords only.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
record_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map record into sub-record (in terms of labels) that is important in the current node-level; mapping should be specified with child_id for
additional filtering
IMPORTANT: SHOULD NOT MODIFY RECORD INTERNALLY, BUT RETURN A NEW RECORD!
'''
#derivable fields:
self.max_depth = len(label_mappings)-1
self.get_labels_of_record = get_labels_of_record
#Build the classification tree.
PRINTER("[MlHierarchical: init]: start of training...")
self.mltree = tree_dfs_builder(frecords,
lambda records, curr_depth, child_id: self.node_builder(records, curr_depth, child_id, classifier, record_mappings),
lambda records, curr_depth, child_id: self.children_splitter(records, curr_depth, child_id, label_mappings), self.max_depth, 0, '')#we start at depth 0
def __init__(self, frecords, classifier, label_mappings, get_labels_of_record):
'''
Constructor.
frecords - generator returning records.
NOTE: if a user wants to manipulate, which codes to consider(e.g. higher or lower level)
it is good to give a specific frecords parameter
classifier - constructor of a multi-label classifier, parametrized by frecords and get_labels_of_record only.
label_mappings - a list equal to depth of a classification tree that is to be constructed. consecutive elements are functions that
map label into sub-label that is important in the current node-level
get_labels_of_record - returns a list of labels of the record
'''
#derivable fields:
self.max_depth = len(label_mappings)-1
self.get_labels_of_record = get_labels_of_record
#Build the classification tree.
PRINTER("[init]: start of training...")
self.mltree = tree_dfs_builder(frecords,
lambda records, curr_depth, child_id: self.node_builder(records, curr_depth, child_id, classifier, label_mappings, get_labels_of_record),
lambda records, curr_depth, child_id: self.children_splitter(records, curr_depth, child_id, label_mappings, get_labels_of_record),
self.max_depth, 0, '', lambda x: True)#we start at depth 0
