Skip to content

Instantly share code, notes, and snippets.

@jcbozonier
jcbozonier / gist:7da348d57a1dcb8acbfe
Last active August 29, 2015 14:15 — forked from anonymous/gist:f77b3da951681d3d2528
TCP server that streams currently tweeting twitter handles of those around NYC for use via Apache Spark. Code HEAVILY borrowed from http://kmkeen.com/socketserver/ Made to work with Spark by Justin Bozonier
import SocketServer, subprocess, sys
from threading import Thread
from TwitterAPI import TwitterAPI
import json, unidecode
api = TwitterAPI(consumer_key, consumer_secret, access_token_key, access_token_secret)
HOST = 'localhost'
PORT = 9999
@jcbozonier
jcbozonier / py.py
Created January 30, 2017 13:38 — forked from anonymous/py.py
Switching from Beta to Dirichlet
hypotheses = [[.8, .2],
[.5,.5],
[.2,.8]]
# Notice how we swapped out the Beta for
# a Dirichlet. The only difference is we
# now pass a list of counts to the pdf
# function. We'll get to why in a bit.
pdf_score = np.array([ss.dirichlet.pdf(hypothesis, [1+1+2, 1+5+3]) for hypothesis in hypotheses])
probabilities = pdf_score/pdf_score.sum()
@jcbozonier
jcbozonier / py.py
Created January 30, 2017 13:39 — forked from anonymous/py.py
Evaluating Walnut and Almond Mix
# The walnut/almond mix ratios we have hypothesized
hypotheses = [[.8, .2],
[.5,.5],
[.2,.8]]
# Evaluate the pdf for each hypothesis
# Note that we only evaluate the hypothesis
# for one nut. If it's 80% we know the
# other must be 20%.
pdf_score = np.array([ss.beta.pdf(hypothesis[0], 1+1, 1+5) for hypothesis in hypotheses])