test_fuzzy_python.py

# -*- coding: utf-8 -*-
"""
Created on Wed Dec  4 11:54:21 2019

@author: rhann
"""

# Goal 1:

# Write a script in Python/Java (other other programming language of interest)
# that a) tries to find the issue we showed in that notebook above
# Add a bit to your script that runs this expression a few (~100) times,
# and prints the result each time
# Finally, add a piece that averages the results and prints the aggregate

# Goal 2:
# Get access to a system that has Docker installed
# (I will look for these for you, as well)
# Run your script in the Docker container and look at if the
# results are different from the "healthy" environment

import numpy as np


def first_function(N):
    step_one = 1/N
    step_two = sum([step_one for _ in range(N)])

    return step_two


def second_function(N, B):
    for _ in range(B):
        M = N ** 0.5
        N = M ** 2

    return N


def u(K, dtype=float):
    K = dtype(K)
    if K == 0:
        return dtype(2)
    elif K == 1:
        return dtype(-4)
    else:
        uminus1 = u(K - 1)
        uminus2 = u(K - 2)
        return 111 - (1130/uminus1) + 3000/(uminus1 * uminus2)


# X(30)
# X(30, dtype=np.float32)
# X(30, dtype=np.float64)
def X(N, dtype=float):
    N = dtype(N)
    if N == 0:
        return dtype(1.5100050721319)
    else:
        xminus1 = X(N - 1)
        return ((3*xminus1**4) - (20*xminus1**3) + (35*xminus1**2) - 24) / \
               ((4*xminus1**3) - (30*xminus1**2) + (70*xminus1 - 50))


def run_x_times(function, x, *args):
    result = []
    for _ in range(x):
        result += [function(*args)]
        sumresult = sum(result)/x
    return sumresult, np.mean(result), np.std(result)


# for every value of k up to 30:
# run u(k) ~ 100 times
# compute std and mean
# print k, mean, and std to screen
for t in [np.float32, np.float64]:
    for _ in range(30):
        print(run_x_times(u, 100, _, t))

# print(run_x_times(third_function, 100000, 100000))
# print(X(30))
# print(u(30))