python(scipy)-- setting an array element with a sequence error

Question

i want to create a program which calculates the average of N random numbers taken from a uniform random number distribution.The program must run for Ν=10,100,1000,10000,100000,1000000 random numbers.Then,i have to plot the mean value as a function of N.

I did this:

from scitools.std import *
import matplotlib.pyplot as plt


N=10

distribution=[]
for i in range(1,7):   
    N*=10
    random_numbers=[random.uniform(0,1,size=N)]   
    distribution.append(random_numbers)


plt.semilogx(array(range(N)),array(distribution).mean())
plt.xlabel('N')
plt.grid(True)
plt.show()

It gives me the error in the title in the line where i do the plot. Also,if there is another ,more pythonic way of doing this i'll appreciate it.

Thank you.

It is in the title:setting an array element with a sequence (Valueerror) — George
– George, Commented Oct 31, 2011 at 16:08
Take the mean of random_numbers before adding it to distribution. — Raymond Hettinger
– Raymond Hettinger, Commented Oct 31, 2011 at 16:09
Thank you all for your answers.I chose the most pythonic.Thanks again! — George
– George, Commented Oct 31, 2011 at 16:20

unutbu · Accepted Answer · 2011-10-31 16:28:34Z

1

import scipy
import matplotlib.pyplot as plt

Ns=[10**i for i in range(1,7)]
means=[scipy.random.uniform(0,1,size=N).mean() for N in Ns]

plt.semilogx(Ns,means)
plt.xlabel('N')
plt.grid(True)
plt.show()

As has already been mentioned, distribution is a list of arrays, with the inner arrays having different shapes. You can't form a numpy array out of such an object by calling np.array(distribution).
At least for plotting purposes, you don't need to save the entire distribution. Just compute and store the means. A succinct way to do that is to use a list comprehension.
Don't use from module import * in scripts. It makes it hard to trace where variables come from. The from module import * was made (mainly) for use in interactive sessions, but generally not recommended for scripts.

edited Oct 31, 2011 at 16:28

answered Oct 31, 2011 at 16:14

unutbu

886k197 gold badges1.9k silver badges1.7k bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

George Over a year ago

Ok,thanks for the tips!I use import * because some times its difficult to know what method goes where.

unutbu Over a year ago

It pays to take the time to sort it out now, and save examples of how you use various functions so you can cut-and-paste later. Doing from scipy import * is particularly dangerous because scipy.any is a different function than Python's builtin any. from scipy import * overwrites Python's builtin any, making it harder to access (and a potential source of bugs).

Raymond Hettinger · Accepted Answer · 2011-10-31 16:29:43Z

1

Compute the mean() before adding the random numbers to the distribution list. Then you have a simple plot of an n-element list. The problem is that you needed a list of scalars, not a list of arrays:

distribution=[]
for i in range(1,7):   
    N*=10
    random_numbers=[random.uniform(0,1,size=N)]   
    distribution.append(array(random_numbers).mean())

edited Oct 31, 2011 at 16:29

answered Oct 31, 2011 at 16:12

Raymond Hettinger

229k67 gold badges405 silver badges504 bronze badges

4 Comments

Sven Marnach Over a year ago

+1. Just for the sake of completeness, I'll add a bit of explanation: The OP's conversion array(distribution) didn't work because it tried to convert a list of one-dimensional arrays of different lengths to a NumPy array, which is not possible. map(numpy.mean, distribution) would also work in the original code.

George Over a year ago

The problem with the above code is that the "mean" method doesn't work because random_numbers is a list.If i make it array,then it gives me error:must have same first dimension

Raymond Hettinger Over a year ago

A list of scalars should be convertible to an array so you can take the mean.

Raymond Hettinger Over a year ago

This works fine: >>> numpy.random.uniform(0,1,10).mean() --> 0.48908919057971734

Blender · Accepted Answer · 2011-10-31 16:10:50Z

0

You seem to be appending a list in a list, like this:

a = [[1], [2], [3]]

Try this code instead:

for i in range(1, 7):   
    N*=10
    distribution.append(random.uniform(0,1,size=N))

@Raymond suggests to take the mean of the sequence:

plt.semilogx(array(range(N)),array(sum(distribution) / float(len(distribution))))

answered Oct 31, 2011 at 16:10

Blender

300k55 gold badges462 silver badges511 bronze badges

Collectives™ on Stack Overflow

python(scipy)-- setting an array element with a sequence error

3 Answers 3

2 Comments

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related