Numpy Exercise 3

Imports


In [11]:
import numpy as np
%matplotlib inline
import matplotlib.pyplot as plt
import seaborn as sns

In [12]:
import antipackage
import github.ellisonbg.misc.vizarray as va


Downloading:  https://raw.githubusercontent.com/ellisonbg/misc/master/vizarray.py
Using existing version:  github.ellisonbg.misc.vizarray

Geometric Brownian motion

Here is a function that produces standard Brownian motion using NumPy. This is also known as a Wiener Process.


In [13]:
def brownian(maxt, n):
    """Return one realization of a Brownian (Wiener) process with n steps and a max time of t."""
    t = np.linspace(0.0,maxt,n)
    h = t[1]-t[0]
    Z = np.random.normal(0.0,1.0,n-1)
    dW = np.sqrt(h)*Z
    W = np.zeros(n)
    W[1:] = dW.cumsum()
    return t, W

Call the brownian function to simulate a Wiener process with 1000 steps and max time of 1.0. Save the results as two arrays t and W.


In [17]:
# YOUR CODE HERE
brownian(1.0,1000)

t,W = np.array(brownian(1.0,1000))

In [18]:
assert isinstance(t, np.ndarray)
assert isinstance(W, np.ndarray)
assert t.dtype==np.dtype(float)
assert W.dtype==np.dtype(float)
assert len(t)==len(W)==1000

Visualize the process using plt.plot with t on the x-axis and W(t) on the y-axis. Label your x and y axes.


In [19]:
plt.plot(t,W)


Out[19]:
[<matplotlib.lines.Line2D at 0x7fbf614017d0>]

In [20]:
assert True # this is for grading

Use np.diff to compute the changes at each step of the motion, dW, and then compute the mean and standard deviation of those differences.


In [25]:
dW = np.diff(W)
print (dW.mean())
print (dW.std())


0.000480726110351
0.0326604764321

In [26]:
assert len(dW)==len(W)-1
assert dW.dtype==np.dtype(float)

Write a function that takes $W(t)$ and converts it to geometric Brownian motion using the equation:

$$ X(t) = X_0 e^{((\mu - \sigma^2/2)t + \sigma W(t))} $$

Use Numpy ufuncs and no loops in your function.


In [45]:
def geo_brownian(t, W, X0, mu, sigma):
    "Return X(t) for geometric brownian motion with drift mu, volatility sigma."""
    a = (((mu - (sigma ** 2))/2)*t) + (sigma * W(t))
    y = (X0)*(np.exp(a))
    return(t,y)

In [46]:
assert True # leave this for grading

Use your function to simulate geometric brownian motion, $X(t)$ for $X_0=1.0$, $\mu=0.5$ and $\sigma=0.3$ with the Wiener process you computed above.

Visualize the process using plt.plot with t on the x-axis and X(t) on the y-axis. Label your x and y axes.


In [ ]:
# YOUR CODE HERE
raise NotImplementedError()

In [ ]:
assert True # leave this for grading