Bollegala

Bollegala

# Check whether for each demo paper, the authors of that paper has a poster to be presented in the same session
import collections
import pandas as pd
import numpy as np

def main():
    posters_df = pd.read_csv('posters.csv')  
    demos_df = pd.read_csv('demos.csv')  

    posters_selected = posters_df[["Session", "Author"]]

Bollegala

property theURLPrefixes : {"https://www.aclweb.org/anthology/"}
property thePDFExtension : ".pdf"


on hasAnyPrefix(theString, thePrefixes)
	repeat with thePrefix in thePrefixes
		if theString starts with thePrefix then return true
	end repeat
	return false
end hasAnyPrefix

Bollegala

from statsmodels.stats.proportion import proportion_confint as binofit
# download statmodels from: http://statsmodels.sourceforge.net/
import sys

acc = float(sys.argv[1])

# 0.01 confidence intervals
easy_interval = binofit(acc * 400, 400, 0.01)
print "0.01 interval = ", easy_interval
strict_interval = binofit(acc * 400, 400, 0.001)

Bollegala

import math
import sys
r = float(sys.argv[1])
num = int(sys.argv[2])
stderr = 1.0 / math.sqrt(num - 3)
delta = 1.96 * stderr
lower = math.tanh(math.atanh(r) - delta)
upper = math.tanh(math.atanh(r) + delta)
print "lower %.6f upper %.6f" % (lower, upper)

Bollegala

"""
This code shows how to perform convolution between two vectors.

Danushka Bollegala.
14-04-2016
"""


import numpy

Bollegala

import numpy
import matplotlib.pyplot as plt
import sys


def bval(D, r, s):
    n = D.shape[0]
    total_r = numpy.sum(D[:,s] ** 2)
    total_s = numpy.sum(D[r,:] ** 2)
    total = numpy.sum(D ** 2)

Bollegala

"""
This script demonstrates the optimization for a oblong quadratic function
using stochastic gradient descent (SGD), SGD with classical momentum (CM),
and SGD with Nestrov's accelerated gradient (NAG).

You will require matplotlib and numpy to run this example.

Danushka Bollegala.
20th June 2015.
"""

Bollegala

"""
This program compares ADAM vs AdaGrad. You can modify the function f and its gradient grad_f 
in the code, run the two algorithms and compare their convergence. For the simple function
f(x1, x2) = (x1 - 2) ** 2 + (x1 + 3) ** 2, (alpha = 0.1 and tolerence 1e-3)
AdaGrad converged at 2023 iterations, whereas ADAM required only 83!
"""

import numpy

Bollegala

from theano import Tensor as T
# If you want to multiply two matrices (matrix product) in theano do the following
T.dot(X,Y)
# This will multiply matrix X (n,m) into the matrix Y(m,k) and produce a matrix (n,k)
# You can use this to compute inner-product between two row vectors as well
T.dot(w,x)
# will give the inner-product wx\T where w(n,) and x(n,)