Module cognet.model

Classes

class model

Facilitate training and constructing Qnet

Init

Expand source code

class model:
    """Facilitate training and constructing Qnet
    """

    def __init__(self):
        """Init
        """
        self.myQnet = None
        self.features = None
        self.immutable_vars = None
        self.mutable_vars = None
        self.data_obj = None

    def fit(self,
            featurenames=None,
            samples=None,
            data_obj=None,
            min_samples_split=2,
            alpha=0.05,
            max_depth=-1,
            max_feats=-1,
            early_stopping=False,
            verbose=0,
            random_state=None,
            njobs=4):
        """fit Quasinet Qnet model

        Args:
          featurenames ([str], optional): names of the model features. Defaults to None.
          samples ([str], optional): 2D array with rows as observations and columns as features. Defaults to None.
          data_obj (obj, optional): Build Qnet directly from data obj without other inputs. Defaults to None.
          njobs (int, optional): Number of jobs used to fit Qnet. Defaults to 2.
        """
        num_None = assert_None([featurenames,samples,data_obj], raise_error=False)
        if num_None == 0:
            raise ValueError("input either samples and features or data object, not both!")
        elif data_obj is not None:
            featurenames, samples=data_obj.Qnet_formatter() # returns the training data
            print(len(samples))
            self.immutable_vars, self.mutable_vars = data_obj.immutable_vars, data_obj.mutable_vars
        elif num_None > 1:
            raise ValueError("input both samples and features or data object!")
        print("training Qnet -------------")
        self.myQnet = Qnet(n_jobs=njobs, feature_names=featurenames,
                           min_samples_split=min_samples_split, alpha=alpha,
                           max_depth=max_depth, max_feats=max_feats,
                           early_stopping=early_stopping,
                           verbose=verbose, random_state=random_state)
        self.myQnet.fit(samples)
        print("Qnet trained --------------")
        self.features = featurenames

    def save(self,
             file_path=None,
             low_mem = False):
        """save qnet

        Args:
          file_path (str, optional): Desired Qnet filename. Defaults to None.
        """
        assert_None([self.myQnet])
        if file_path is None:
            file_path = 'tmp_Qnet.joblib'
        save_qnet(self.myQnet, file_path, low_mem=low_mem)
    
    def load(self,
             file_path,
             VERBOSE=False):
        """load Qnet from file

        Args:
          file_path (str): path to Qnet savefile
          VERBOSE (bool): boolean to turn on verbose

        Returns:
          [Qnet]: Qnet object
        """
        if VERBOSE:
            print("loading..")
        # check if reading from url or local directory
        if "https" in file_path or "www." in file_path:
            url_file = BytesIO(requests.get(file_path).content)
            self.myQnet = load_qnet(url_file)
        else:
            self.myQnet = load_qnet(file_path)
        self.features = self.myQnet.feature_names
        if VERBOSE:
            print("done")
        return self.myQnet

    def export_dot(self,
                   filename,
                   index=[3],
                   path='',
                   generate_trees=False,
                   threshold=0.2):
        """export Qnet trees

        Args:
          filename (str): Desired tree savefile
          index (list, optional): list of indices to generate trees. Defaults to [3].
          path (str, optional): Desired tree savefile path. Defaults to ''.
          generate_trees (bool, optional): Whether or not to generate individual trees. 
                                             Defaults to False, or to generate individual trees.
          threshold (float, optional): Numeric cutoff for edge weights. If the edge weights exceed 
                                         this cutoff, then we include it into the graph. Defaults to 0.2.
        """
        if not generate_trees:
            export_qnet_graph(self.myQnet, 
                              threshold, path+filename)
        else:
            for i in index:
                export_qnet_tree(self.myQnet,
                                 i, filename)

Methods

def export_dot(self, filename, index=[3], path='', generate_trees=False, threshold=0.2)

export Qnet trees

Args

filename : str

Desired tree savefile

index : list, optional

list of indices to generate trees. Defaults to [3].

path : str, optional

Desired tree savefile path. Defaults to ''.

generate_trees : bool, optional

Whether or not to generate individual trees. Defaults to False, or to generate individual trees.

threshold : float, optional

Numeric cutoff for edge weights. If the edge weights exceed this cutoff, then we include it into the graph. Defaults to 0.2.

Expand source code

def export_dot(self,
               filename,
               index=[3],
               path='',
               generate_trees=False,
               threshold=0.2):
    """export Qnet trees

    Args:
      filename (str): Desired tree savefile
      index (list, optional): list of indices to generate trees. Defaults to [3].
      path (str, optional): Desired tree savefile path. Defaults to ''.
      generate_trees (bool, optional): Whether or not to generate individual trees. 
                                         Defaults to False, or to generate individual trees.
      threshold (float, optional): Numeric cutoff for edge weights. If the edge weights exceed 
                                     this cutoff, then we include it into the graph. Defaults to 0.2.
    """
    if not generate_trees:
        export_qnet_graph(self.myQnet, 
                          threshold, path+filename)
    else:
        for i in index:
            export_qnet_tree(self.myQnet,
                             i, filename)

def fit(self, featurenames=None, samples=None, data_obj=None, min_samples_split=2, alpha=0.05, max_depth=-1, max_feats=-1, early_stopping=False, verbose=0, random_state=None, njobs=4)

fit Quasinet Qnet model

Args

featurenames : [str], optional

names of the model features. Defaults to None.

samples : [str], optional

2D array with rows as observations and columns as features. Defaults to None.

data_obj : obj, optional

Build Qnet directly from data obj without other inputs. Defaults to None.

njobs : int, optional

Number of jobs used to fit Qnet. Defaults to 2.

Expand source code

def fit(self,
        featurenames=None,
        samples=None,
        data_obj=None,
        min_samples_split=2,
        alpha=0.05,
        max_depth=-1,
        max_feats=-1,
        early_stopping=False,
        verbose=0,
        random_state=None,
        njobs=4):
    """fit Quasinet Qnet model

    Args:
      featurenames ([str], optional): names of the model features. Defaults to None.
      samples ([str], optional): 2D array with rows as observations and columns as features. Defaults to None.
      data_obj (obj, optional): Build Qnet directly from data obj without other inputs. Defaults to None.
      njobs (int, optional): Number of jobs used to fit Qnet. Defaults to 2.
    """
    num_None = assert_None([featurenames,samples,data_obj], raise_error=False)
    if num_None == 0:
        raise ValueError("input either samples and features or data object, not both!")
    elif data_obj is not None:
        featurenames, samples=data_obj.Qnet_formatter() # returns the training data
        print(len(samples))
        self.immutable_vars, self.mutable_vars = data_obj.immutable_vars, data_obj.mutable_vars
    elif num_None > 1:
        raise ValueError("input both samples and features or data object!")
    print("training Qnet -------------")
    self.myQnet = Qnet(n_jobs=njobs, feature_names=featurenames,
                       min_samples_split=min_samples_split, alpha=alpha,
                       max_depth=max_depth, max_feats=max_feats,
                       early_stopping=early_stopping,
                       verbose=verbose, random_state=random_state)
    self.myQnet.fit(samples)
    print("Qnet trained --------------")
    self.features = featurenames

def load(self, file_path, VERBOSE=False)

load Qnet from file

Args

file_path : str

path to Qnet savefile

VERBOSE : bool

boolean to turn on verbose

Returns

[Qnet]

Qnet object

Expand source code

def load(self,
         file_path,
         VERBOSE=False):
    """load Qnet from file

    Args:
      file_path (str): path to Qnet savefile
      VERBOSE (bool): boolean to turn on verbose

    Returns:
      [Qnet]: Qnet object
    """
    if VERBOSE:
        print("loading..")
    # check if reading from url or local directory
    if "https" in file_path or "www." in file_path:
        url_file = BytesIO(requests.get(file_path).content)
        self.myQnet = load_qnet(url_file)
    else:
        self.myQnet = load_qnet(file_path)
    self.features = self.myQnet.feature_names
    if VERBOSE:
        print("done")
    return self.myQnet

def save(self, file_path=None, low_mem=False)

save qnet

Args

file_path : str, optional

Desired Qnet filename. Defaults to None.

Expand source code

def save(self,
         file_path=None,
         low_mem = False):
    """save qnet

    Args:
      file_path (str, optional): Desired Qnet filename. Defaults to None.
    """
    assert_None([self.myQnet])
    if file_path is None:
        file_path = 'tmp_Qnet.joblib'
    save_qnet(self.myQnet, file_path, low_mem=low_mem)