def __init__(self, dist_dir):
Pmf.__init__(self)
self.dist_dir = dist_dir
self.pmf = None
self.infs_data = {}
self.json_data = {}
self.dist_file = ''
def __init__(self, hypos):
"""Initialize self.
hypos: sequence of string bowl IDs
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
def __init__(self, hypos):
"""Initialize the distribution.
hypos: sequence of hypotheses
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
def __init__(self, hypos):
"""Initialize self.
hypos: sequence of string bowl IDs
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
def __init__(self, hypos, alpha=1.0):
Pmf.__init__(self)
'''
父类对prior的概率分布采用的是平均分布(uniform distribution)
此处改写prior的概率分布为幂律(power law distribution)
'''
for hypo in hypos:
self.Set(hypo, hypo**(-alpha))
self.Normalize()
def __init__(self, bowl1, bowl2):
"""Constructor"""
self.bowl1 = bowl1
self.bowl2 = bowl2
Pmf.__init__(self)
self.Set(bowl1.bowl_name, 1)
self.Set(bowl2.bowl_name, 1)
self.Normalize()
self.set_mixes()
def __init__(self, hypos):
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
self.mixes = {
'Bowl 1':dict(vanilla=0.75, chocolate=0.25),
'Bowl 2':dict(vanilla=0.5, chocolate=0.5)
}
def __init__(self, hypos):
"""Initialize the distribution.
hypos: sequence of hypotheses
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
def __init__(self, hypos, bowls):
"""Initialize self.
hypos: sequence of string bowl IDs
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
self.bowls = dict()
for bowl in bowls.keys():
self.bowls[bowl] = Bowl(bowls[bowl])
def __init__(self, hypos):
"""Initialize self.
hypos: sequence of string bowl IDs
"""
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
bowl1 = Bowl(vanilla=30, chocolate=10)
bowl2 = Bowl(vanilla=20, chocolate=20)
self.bowls = {'Bowl 1': bowl1, 'Bowl 2': bowl2}
def MakePmf(self, steps=101, label=None):
"""Returns a Pmf of this distribution.
Note: Normally, we just evaluate the PDF at a sequence
of points and treat the probability density as a probability
mass.
But if alpha or beta is less than one, we have to be
more careful because the PDF goes to infinity at x=0
and x=1. In that case we evaluate the CDF and compute
differences.
The result is a little funny, because the values at 0 and 1
are not symmetric. Nevertheless, it is a reasonable discrete
model of the continuous distribution, and behaves well as
the number of values increases.
"""
if label is None and self.label is not None:
label = self.label
if self.alpha < 1 or self.beta < 1:
cdf = self.MakeCdf()
pmf = cdf.MakePmf()
return pmf
xs = [i / (steps - 1) for i in range(steps)]
probs = [self.EvalPdf(x) for x in xs]
pmf = Pmf(dict(zip(xs, probs)), label=label)
return pmf
def main():
pmf = Pmf()
pmf.Set('Bowl 1', 0.5)
pmf.Set('Bowl 2', 0.5)
pmf.Mult('Bowl 1', 0.75)
pmf.Mult('Bowl 2', 0.5)
pmf.Normalize()
print(pmf.Prob('Bowl 1'))
print(pmf.Prob('Bowl 2'))
# -*- coding: utf-8 -*-
"""
Created on Tue Apr 17 12:08:20 2018
@author: QJ
"""
###Think Bayes
from thinkbayes2 import Pmf
pmf = Pmf()
for x in [1, 2, 3, 4, 5, 6]:
pmf.Set(x, 1 / 6.0)
pmf.Set('Bowl1', 0.5)
pmf.Set('Bowl2', 0.5)
pmf.Mult('Bowl1', 0.75)
pmf.Mult('Bowl2', 0.5)
print pmf.Prob('Bowl 1')
print(pmf)
# http://www.greenteapress.com/thinkbayes/thinkbayes.pdf
# http://thinkbayes.com/thinkbayes.py
# python -m pip install scipy numpy matplotlib pandas
from thinkbayes2 import Pmf
pmf = Pmf()
pmf.Set('tazon1', 0.5)
pmf.Set('tazon2', 0.5)
pmf.Mult('tazon1', 0.75)
pmf.Mult('tazon2', 0.5)
pmf.Normalize()
print(pmf.Prob('tazon1'))
def __init__(self, hypos, priors):
Pmf.__init__(self)
'''Set the priors (for each H, the corrisponding p(H)'''
for i in range(len(hypos)):
self.Set(hypos[i],priors[i])
self.Normalize()
def __init__(self, hypos, alpha=None):
Pmf.__init__(self, hypos)
if alpha is not None:
for hypo in hypos:
self.Set(hypo, hypo**(-alpha))
self.Normalize()
def __init__(self, sides):
Pmf.__init__(self)
for x in range(1, sides + 1):
self.Set(x, 1)
self.Normalize()
#!/usr/env python
import sys
sys.path.append("../lib/ThinkBayes2/code/")
from thinkbayes2 import Pmf
pmf = Pmf()
#Prior Dist
pmf.Set('Bowl1', 0.5)
pmf.Set('Bowl2', 0.5)
print(pmf)
#Update based on new data
pmf.Mult('Bowl1', 0.75)
pmf.Mult('Bowl2', 0.5)
print(pmf)
#Hypotheses are mutally exclusive and collectively exhaustive, we can renormalize!
pmf.Normalize()
print(pmf) #posterior distribution
#book covers a more complicated implementation, but it is the same as Monty
print("Finished")
from __future__ import print_function, division
import sys
sys.path.append("../lib/ThinkBayes2/code/")
from thinkbayes2 import Suite, Pmf, SampleSum, MakeMixture
import thinkplot
from simulationDD02 import Die
pmf_dice = Pmf()
pmf_dice.Set(Die(4), 2)
pmf_dice.Set(Die(6), 3)
pmf_dice.Set(Die(8), 2)
pmf_dice.Set(Die(12), 1)
pmf_dice.Set(Die(20), 1)
pmf_dice.Normalize()
print(pmf_dice)
print("#################################################")
mix = Pmf()
for die, weight in pmf_dice.Items():
for outcome, prob in die.Items():
mix.Incr(outcome, weight * prob)
#Shorthand for above
#mix = MakeMixture(pmf_dice)
print(mix)
thinkplot.Hist(mix)
thinkplot.Save(root='bar',
xlabel='Mixture over a set of dice',
ylabel='Probability',
formats=['pdf'])
from thinkbayes2 import Pmf
if __name__ == '__main__':
txt = []
with open('/Users/glenn/math/stats/ThinkBayes2/data/ch02_02_text_data.txt'
) as f:
txt = f.read().split()
print "txt = ", txt
pmf = Pmf()
for word in txt:
pmf.Incr(word, 1)
pmf.Normalize()
pmf.Print()
from thinkbayes2 import Pmf
import thinkplot
d6 = Pmf()
for x in [1, 2, 3, 4, 5, 6]:
d6[x] = 1
d6.Normalize()
print(d6)
twice = d6 + d6
# If the sum of two dice is greater than 3, then we update the dictionary
twice[2] = 0
twice[3] = 0
twice.Normalize()
print(twice)
thinkplot.Hist(twice)
thinkplot.show()
from thinkbayes2 import Pmf
if __name__ == '__main__':
pmf = Pmf()
for x in range(1, 7):
pmf.Set(x, 1 / 6.0)
pmf.Print()
#!/usr/env python
import sys
sys.path.append("../lib/ThinkBayes2/code/")
from thinkbayes2 import Pmf
pmf = Pmf()
for x in [1, 2, 3, 4, 5, 6]:
pmf.Set(x, 1 / 6.0)
print(pmf)
words = Pmf()
f = open('alice.txt', 'r')
for line in f:
for word in line.split():
words.Incr(word.strip(), 1)
#print(words)
words.Normalize()
print(words)
print("Finished")
from thinkbayes2 import Pmf, Suite
import thinkplot
# Probability of each dice given that when rolled the output is 6
dice = Pmf(['4-sided', '6-sided', '8-sided', '12-sided'])
dice['4-sided'] *= 0
dice['6-sided'] *= 1 / 6
dice['8-sided'] *= 1 / 8
dice['12-sided'] *= 1 / 12
dice.Normalize()
print(dice)
suite = Suite([4, 6, 8, 12])
suite[4] *= 0
suite[6] *= 1 / 6
suite[8] *= 1 / 8
suite[12] *= 1 / 12
suite.Normalize()
print(suite)
class Dice(Suite):
# hypo is the number if sides in the die
# dat is the outcome
def Likelihood(self, data, hypo):
return 0 if data > hypo else 1 / hypo
class InfulsContribution(Pmf):
def __init__(self, dist_dir):
Pmf.__init__(self)
self.dist_dir = dist_dir
self.pmf = None
self.infs_data = {}
self.json_data = {}
self.dist_file = ''
def prior(self, json_data):
self.pmf = Pmf()
self.posterior(json_data)
return self
def posterior(self, json_data):
self.set_json_data(json_data)
self.update_infs_data()
self.update()
self.set_dist_file(json_data, self.dist_dir)
self.output()
return self
def output(self):
pmf = self.pmf
pmf.Normalize()
dist_data = self.make_result(pmf)
self.save_file(dist_data)
return self
# main nethods
# set jsondata as dict to self.jsondata
def set_json_data(self, json_data):
with open(json_data, 'rb') as f:
data = json.load(f)
self.json_data = data
# update self.infs_data
def update_infs_data(self):
# print('sel', self.json_data)
datas = self.json_data['analythics_data']
click_sum = self.json_data['meta_data']['click_sum']
# infs_data = 0
# for d in datas:
# infs_data += int(datas[d]['prof_clicks'])
for d in datas:
if not self.infs_data.get(d):
self.infs_data[d] = {'click_sum': 0, 'prof_clicks': 0}
self.infs_data[d]['click_sum'] += click_sum
self.infs_data[d]['prof_clicks'] += datas[d]['prof_clicks']
# set self dist_file
def set_dist_file(self, src_file, dist_dir):
dist_file = ''
a = re.search(r'\d{4}-\d{2}-\d{2}', src_file)
b = re.search(r'\d{8}', src_file)
if a: dist_file = a.group()
if b: dist_file = b.group()
dist_file = os.path.join(dist_dir, f'result{dist_file}.json')
self.dist_file = dist_file
# LiKelihood methods
def liKelihood(self, infl):
return (self.infs_data[infl]['prof_clicks'] /
self.infs_data[infl]['click_sum'])
# update datas
def update(self):
data = self.json_data['analythics_data']
click_sum = self.json_data['meta_data']['click_sum']
print('d', data)
# data = {'@yoshiki_ruby': data['@yoshiki_ruby']}
[self.pmf.Incr(infl, self.liKelihood(infl)) for infl in data]
# create result json
def make_result(self, pmf):
d = pmf.Values()
res_d = {k: pmf.Prob(k) for k in d}
res_d = sorted(res_d.items(), key=lambda x: x[1], reverse=True)
# print(res_d)
return {
v[0]: {
"rating": v[1],
"prof_clicks": self.infs_data[v[0]]['prof_clicks'],
"click_sum": self.infs_data[v[0]]['click_sum'],
}
for v in res_d
}
# output json
def save_file(self, json_file):
with open(self.dist_file, 'w') as f:
json.dump(json_file, f, indent=2)
"""This file contains code for use with "Think Bayes",
by Allen B. Downey, available from greenteapress.com
Copyright 2012 Allen B. Downey
License: GNU GPLv3 http://www.gnu.org/licenses/gpl.html
"""
from __future__ import print_function, division
from thinkbayes2 import Pmf
pmf = Pmf()
pmf.Set('Bowl 1', 0.5)
pmf.Set('Bowl 2', 0.5)
pmf.Mult('Bowl 1', 0.75)
pmf.Mult('Bowl 2', 0.5)
pmf.Normalize()
print(pmf.prob('Bowl 1'))
import sys
sys.path.insert(0, '/Users/carol/python/ThinkBayes2/thinkbayes2/')
import numpy as np
import matplotlib.pyplot as plt
from thinkbayes2 import Pmf, Suite, CredibleInterval, Beta
# PMF for 6-sided die
pmf = Pmf()
for x in [1, 2, 3, 4, 5, 6]:
pmf.Set(x, 1 / 6)
print(pmf)
# How to build up a pmf from a list of strings
pmf2 = Pmf()
for word in ['a', 'in', 'or', 'to', 'a', 'me', 'in']:
pmf2.Incr(word, 1)
pmf2.Normalize()
print(pmf2)
print("Probability of letter a:", pmf2.Prob(
'a')) # Typo p12 print pmf.Prob('the') should read print(pmf.Prob('the'))
# PMF for the Cookie problem
pmf = Pmf()
# Prior:
pmf.Set("Bowl 1", 0.5)
pmf.Set("Bowl 2", 0.5)
# Posterior:
# First multiply prior by likelihood
pmf.Mult("Bowl 1", 0.75)
def prior(self, json_data):
self.pmf = Pmf()
self.posterior(json_data)
return self
def __init__(self, hypos):
Pmf.__init__(self)
for hypo in hypos:
self.Set(hypo, 1)
self.Normalize()
from thinkbayes2 import Pmf # Cookie Problem # Prior 50%-50% cookie = Pmf(['Bowl 1', 'Bowl 2']) # p (Vanilla|Bowl1) = 30/40 cookie['Bowl 1'] *= 0.75 # p (Vanilla|Bowl2) = 20/40 cookie['Bowl 2'] *= 0.5 # Normalize, return values is p(D) cookie.Normalize() # Posteriors print(cookie) # Suppose we put the first cookie back, stir, choose again from the same bowl, and # get a chocolate cookie # p (Chocolate|Bowl1) = 10/40 cookie['Bowl 1'] *= 0.25 # p (Chocolate|Bowl2) = 20/40 cookie['Bowl 2'] *= 0.5 cookie.Normalize() print(cookie)
