"""Randomized Causation Coefficient Model.
Author : David Lopez-Paz
Ref : Lopez-Paz, David and Muandet, Krikamol and Schölkopf, Bernhard and Tolstikhin, Ilya O,
"Towards a Learning Theory of Cause-Effect Inference", ICML 2015.
.. MIT License
..
.. Copyright (c) 2018 Diviyan Kalainathan
..
.. Permission is hereby granted, free of charge, to any person obtaining a copy
.. of this software and associated documentation files (the "Software"), to deal
.. in the Software without restriction, including without limitation the rights
.. to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
.. copies of the Software, and to permit persons to whom the Software is
.. furnished to do so, subject to the following conditions:
..
.. The above copyright notice and this permission notice shall be included in all
.. copies or substantial portions of the Software.
..
.. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
.. IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
.. FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
.. AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
.. LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
.. OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
.. SOFTWARE.
"""
from sklearn.preprocessing import scale
from sklearn.ensemble import RandomForestClassifier as CLF
from ...utils.Settings import SETTINGS
import pandas
import numpy as np
from .model import PairwiseModel
[docs]class RCC(PairwiseModel):
"""Randomized Causation Coefficient model. 2nd approach in the Fast
Causation challenge.
**Description:** The Randomized causation coefficient (RCC) relies on the
projection of the empirical distributions into a RKHS using random cosine
embeddings, then classfies the pairs using a random forest based on those
features.
**Data Type:** Continuous, Categorical, Mixed
**Assumptions:** This method needs a substantial amount of labelled causal
pairs to train itself. Its final performance depends on the training set
used.
Args:
rand_coeff (int): number of randomized coefficients
nb_estimators (int): number of estimators
nb_min_leaves (int): number of min samples leaves of the estimator
max_depth (): (optional) max depth of the model
s (float): scaling
njobs (int): number of jobs to be run on parallel (defaults to ``cdt.SETTINGS.NJOBS``)
verbose (bool): verbosity (defaults to ``cdt.SETTINGS.verbose``)
.. note::
Ref : Lopez-Paz, David and Muandet, Krikamol and Schölkopf, Bernhard and Tolstikhin, Ilya O,
"Towards a Learning Theory of Cause-Effect Inference", ICML 2015.
Example:
>>> from cdt.causality.pairwise import RCC
>>> import networkx as nx
>>> import matplotlib.pyplot as plt
>>> from cdt.data import load_dataset
>>> from sklearn.model_selection import train_test_split
>>> data, labels = load_dataset('tuebingen')
>>> X_tr, X_te, y_tr, y_te = train_test_split(data, labels, train_size=.5)
>>>
>>> obj = RCC()
>>> obj.fit(X_tr, y_tr)
>>> # This example uses the predict() method
>>> output = obj.predict(X_te)
>>>
>>> # This example uses the orient_graph() method. The dataset used
>>> # can be loaded using the cdt.data module
>>> data, graph = load_dataset('sachs')
>>> output = obj.orient_graph(data, nx.DiGraph(graph))
>>>
>>> # To view the directed graph run the following command
>>> nx.draw_networkx(output, font_size=8)
>>> plt.show()
"""
def __init__(self, rand_coeff=333, nb_estimators=500, nb_min_leaves=20,
max_depth=None, s=10, njobs=None, verbose=None):
"""Initialize the model w/ its parameters.
"""
np.random.seed(0)
self.K = rand_coeff
self.E = nb_estimators
self.L = nb_min_leaves
self.njobs, self.verbose = SETTINGS.get_default(('njobs', njobs), ('verbose', verbose))
self.max_depth = max_depth
self.W = np.hstack((s * np.random.randn(self.K, 2),
2 * np.pi * np.random.rand(self.K, 1)))
self.W2 = np.hstack((s * np.random.randn(self.K, 1),
2 * np.pi * np.random.rand(self.K, 1)))
self.clf = None
[docs] def featurize_row(self, x, y):
""" Projects the causal pair to the RKHS using the sampled kernel approximation.
Args:
x (np.ndarray): Variable 1
y (np.ndarray): Variable 2
Returns:
np.ndarray: projected empirical distributions into a single fixed-size vector.
"""
x = x.ravel()
y = y.ravel()
b = np.ones(x.shape)
dx = np.cos(np.dot(self.W2, np.vstack((x, b)))).mean(1)
dy = np.cos(np.dot(self.W2, np.vstack((y, b)))).mean(1)
if(sum(dx) > sum(dy)):
return np.hstack((dx, dy,
np.cos(np.dot(self.W, np.vstack((x, y, b)))).mean(1)))
else:
return np.hstack((dx, dy,
np.cos(np.dot(self.W, np.vstack((y, x, b)))).mean(1)))
[docs] def fit(self, x, y):
"""Train the model.
Args:
x_tr (pd.DataFrame): CEPC format dataframe containing the pairs
y_tr (pd.DataFrame or np.ndarray): labels associated to the pairs
"""
train = np.vstack((np.array([self.featurize_row(row.iloc[0],
row.iloc[1]) for idx, row in x.iterrows()]),
np.array([self.featurize_row(row.iloc[1],
row.iloc[0]) for idx, row in x.iterrows()])))
labels = np.vstack((y, -y)).ravel()
verbose = 1 if self.verbose else 0
self.clf = CLF(verbose=verbose,
min_samples_leaf=self.L,
n_estimators=self.E,
max_depth=self.max_depth,
n_jobs=self.njobs).fit(train, labels)
[docs] def predict_proba(self, dataset, **kwargs):
""" Predict the causal score using a trained RCC model
Args:
dataset (tuple): Couple of np.ndarray variables to classify
Returns:
float: Causation score (Value : 1 if a->b and -1 if b->a)
"""
if self.clf is None:
raise ValueError("Model has to be trained before making predictions.")
x, y = dataset
input_ = self.featurize_row(x, y).reshape((1, -1))
return self.clf.predict(input_)