[Kaggle] *Representing and Estimating Bayesian network

If you are using virtualenv or other native package management, you may need to run these commands:

pip install pandas matplotlib graphviz

To render the generated DOT source code, you also need to install Graphviz download page.

Link: https://graphviz.readthedocs.io/en/stable/examples.html

很有意思的一个库，画圈圈，画线线。

 

# Make division default to floating-point, saving confusion
from __future__ import division
from __future__ import print_function

# Necessary libraries
import numpy as np
import pandas as pd
import matplotlib.pyplot as pl
from random import random

# combinatorics
from itertools import product, combinations
# ordered dictionaries are useful for keeping ordered sets of varibles
from collections import OrderedDict as odict
#visualise our graph
from graphviz import Digraph　　# <-- interesting

# Put the graphs where we can see them
%matplotlib inline

# easier debugging display
pd.set_option('display.multi_sparse', False)
from pprint import pprint

 

X	Y	Z	p(Z | X,Y)
0	0	0	0
0	0	1	1
0	1	0	1/3
0	1	1	2/3
1	0	0	1/3
1	0	1	2/3
1	1	0	2/5
1	1	1	3/5

example_prob_table = odict([
    ((0, 0,), [0.0, 1]),
    ((0, 1,), [1/3, 2/3]),
    ((1, 0,), [1/3, 2/3]),
    ((1, 1,), [2/5, 3/5]),
])

 

 

def p_cond(prob_table, this_val, *parents):
    """
    argument 
    `prob_table`, a dictionary of conditional probabilities,
    `this_val` (e.g., 1) the value of the variable we wish to find the probability for
    `parents` (e.g. Y=1, Z=1) the parent values that we wish to set.
    You must use a consistent order for the parents.
    
    Returns p(X=this_val|parents(X)=parents)
    """

    prob_array = np.asarray(prob_table[parents])
    
    if this_val<prob_array.size:
        return prob_array[this_val]
    elif this_val == prob_array.size:
        return 1.0 - prob_array.sum()
    else:
        return 0.0

 

print('p(Z=1|X=1,Y=1)={}'.format( p_cond(example_prob_table, 1, 1, 1) ))

 

 

 

 

 

example_prob_table = odict([    ((0, 0,), [0.0, 1]),    ((0, 1,), [1/3, 2/3]),    ((1, 0,), [1/3, 2/3]),    ((1, 1,), [2/5, 3/5]),])