def __init__(self, domain, jtreepy, _lambda=0.2):
"""
Using linear programming method to find a less noise variance.
param:
domain: the domain of the given data
(note) the order of columns in domain should be same with the original
jtree: the structure of junction tree
_lambda: the balance number.
TODO:
1. Move jt_rep to Junction Tree Module
2. Move Different Operator to a linear algebra package
"""
self.LOG = Base.get_logger("CliqueMerge")
self.domain = domain
self.node_card = [len(vals) for vals in domain.values()]
self._lambda = float(_lambda)
self.max_iter = 20
self.jtree = jtreepy
self.nodes_num = len(self.node_card)
self.cliques_num = len(jtreepy)
self.cnum = range(2, len(jtreepy) + 1) if len(jtreepy) >= 2 else [1]
self.jtree_in_node_index = [
self.find_subset_index(clique) for clique in jtreepy
]
def __init__(
self,
data = None,
edges = None,
noise_flag = True,
white_list = [],
eps1_val = c.EPSILON_1,
cramer = 0.2):
"""
__init__
Input:
1. DataUtils.Data
Procedure
1. Convert the given data frame to dataframe in R
2. Convert the given Domain(in python dict) to ListVector
3. Instantial the attributes dependency.
"""
self.LOG = Base.get_logger("DepGraph")
self.noise_flag = noise_flag
self.eps1_val = eps1_val
self.cramer = cramer
self.data = data
if data is None:
self.edges = edges
else:
self.edges = self._run()
self.white_list = white_list
def __init__(self, sensitive_data):
""" Import the original data and initialize the utility measurement object
Parameters
----------
sensitive_data: string
The path to the original data.
"""
self.LOG = Base.get_logger("UserQuery")
sensitive = DataUtils(file_path=sensitive_data)
self.sensitive_df = sensitive.get_pandas_df()
def __init__(self, sensitive_data):
""" Import the original data and initialize the utility measurement object
Parameters
----------
sensitive_data: string
The path to the original data.
"""
self.LOG = Base.get_logger("UserQuery")
sensitive = DataUtils(file_path = sensitive_data)
self.sensitive_df = sensitive.get_pandas_df()
def __init__(
self,
file_path = None,
selected_attrs = None,
pandas_df = None,
valbin_maps = None,
names=None,
specified_c_domain = None,
chunk_size = -1,
date_format = None
):
""" Loading data to warpped python object
Parameters
----------
file_path: string
The path of original data
selected_attrs: dict
{
"A":"C",
"B":"D",
...
}
pandas_df: Pandas dataframe
Initialize with a pandas dataframe(TODO: deprecated)
valbin_maps: dict
A mapping of original values with coarse value
names: list, experiment
A list to specifiy the attributes' names when the input file has no header
specified_c_domain: dict, experiment
A mapping of continuous type attributes with the specified edges in coarse
"""
self.LOG = Base.get_logger("DataUtils")
self.valbin_maps = dict() if valbin_maps is None else valbin_maps
self.chunk_size = chunk_size
if chunk_size > 0:
self.dataframe = self._loading_chunk(file_path, pandas_df, names)
else:
self.dataframe = self._loading(file_path, pandas_df, names)
if selected_attrs is not None:
self.selected_attrs = selected_attrs
# the 'selected_attrs' is ordered
self.dataframe = self.dataframe[selected_attrs.keys()]
self.preview_count = 5
self.specified_c_domain = specified_c_domain
self.date_format = date_format
def __init__(self,
file_path=None,
selected_attrs=None,
pandas_df=None,
valbin_maps=None,
names=None,
specified_c_domain=None,
chunk_size=-1,
date_format=None):
""" Loading data to warpped python object
Parameters
----------
file_path: string
The path of original data
selected_attrs: dict
{
"A":"C",
"B":"D",
...
}
pandas_df: Pandas dataframe
Initialize with a pandas dataframe(TODO: deprecated)
valbin_maps: dict
A mapping of original values with coarse value
names: list, experiment
A list to specifiy the attributes' names when the input file has no header
specified_c_domain: dict, experiment
A mapping of continuous type attributes with the specified edges in coarse
"""
self.LOG = Base.get_logger("DataUtils")
self.valbin_maps = dict() if valbin_maps is None else valbin_maps
self.chunk_size = chunk_size
if chunk_size > 0:
self.dataframe = self._loading_chunk(file_path, pandas_df, names)
else:
self.dataframe = self._loading(file_path, pandas_df, names)
if selected_attrs is not None:
self.selected_attrs = selected_attrs
# the 'selected_attrs' is ordered
self.dataframe = self.dataframe[selected_attrs.keys()]
self.preview_count = 5
self.specified_c_domain = specified_c_domain
self.date_format = date_format
def __init__(self,
data,
jtree_path,
domain,
cluster,
histogramdds,
epsilon = 0.0
):
"""
Initialize the inference class.
TODO: 1. refactor, the data_path, edges, nodes, domain
are temporary to be here.
param
data: the pandas dataframe
TODO: Because the DPTable algorithm construct lots of attributes when reading data,
to using memory cache, one should refector the inference step of DPTable.
param
domain: data information with format in dictionary
{
"A":[1,2,3,4,5],
"B":[2,3,4,5,6]
}
param
cluster: the merged cluster structure
param
epsilon: the privacy budget
"""
self.LOG = Base.get_logger("Inference")
self.data = data
self.data_size = data.get_count()
self.epsilon = epsilon
self.rdomain = self.convert2rdomain(domain)
sorted_internal = lambda ls2: [sorted(ls) for ls in ls2]
self.cluster = sorted_internal(cluster)
self.jtree_path = jtree_path
self.histogramdds = histogramdds
def __init__(self, domain, jtreepy, _lambda=0.2):
"""
Using linear programming method to find a less noise variance.
param:
domain: the domain of the given data
(note) the order of columns in domain should be same with the original
jtree: the structure of junction tree
_lambda: the balance number.
TODO:
1. Move jt_rep to Junction Tree Module
2. Move Different Operator to a linear algebra package
"""
self.LOG = Base.get_logger("CliqueMerge")
self.domain = domain
self.node_card = [len(vals) for vals in domain.values()]
self._lambda = float(_lambda)
self.max_iter = 20
self.jtree = jtreepy
self.nodes_num = len(self.node_card)
self.cliques_num = len(jtreepy)
self.cnum = range(2, len(jtreepy)+1) if len(jtreepy) >=2 else [1]
self.jtree_in_node_index = [self.find_subset_index(clique) for clique in jtreepy]
def __init__(self, edges, nodes, jtree_path=None):
edges = self.convert2rlistofvector(edges)
self.LOG = Base.get_logger("JunctionTree")
self.jtree = self._build_jtree(edges, nodes, jtree_path)
def __init__(self, edges, nodes, jtree_path = None):
edges = self.convert2rlistofvector(edges)
self.LOG = Base.get_logger("JunctionTree")
self.jtree = self._build_jtree(edges, nodes, jtree_path)
