Get average between consecutive pairs of numpy array
Question:
Say I have a numpy array like this
[1,2,3,4,5]
I want to generate an array that is the equal to the average of consecutive elements
[1.5,2.5,3.5,4.5]
Is there any efficient way to do this outside of just iterating through?
I’m not really sure what to do because reshaping doesn’t really work and I’m trying to get a smaller array.
Answers:
Using NumPy’s vectorized operation (highly-optimized functions for performing mathematical operations) in the code below.
arr[:-1] contains all but the last element of arr
arr[1:] contains all but the first element of arr
import numpy as np
arr = np.array([1, 2, 3, 4, 5, 6, 7, 8])
avg_arr = (arr[:-1] + arr[1:]) / 2.0
print(avg_arr)
You can use np.convolve to implement two-element averaging as the convolution of the input vector with the vector [0.5, 0.5]:
>>> a = [1, 2, 3, 4, 5]
>>> np.convolve(a, [0.5, 0.5], "valid")
array([1.5, 2.5, 3.5, 4.5])
np.convolve is a highly optimized routine, and will delegate to different implementations depending on what it estimates will be fastest for the given inputs. For small vectors like [0.5, 0.5] it will likely just compute the result directly (0.5*a[n] + 0.5*[n+1] for each element), but for large inputs it can leverage FFT-based convolution algorithms to compute the result with significantly fewer operations than the direct method.
Say I have a numpy array like this
[1,2,3,4,5]
I want to generate an array that is the equal to the average of consecutive elements
[1.5,2.5,3.5,4.5]
Is there any efficient way to do this outside of just iterating through?
I’m not really sure what to do because reshaping doesn’t really work and I’m trying to get a smaller array.
Using NumPy’s vectorized operation (highly-optimized functions for performing mathematical operations) in the code below.
arr[:-1] contains all but the last element of arr
arr[1:] contains all but the first element of arr
import numpy as np
arr = np.array([1, 2, 3, 4, 5, 6, 7, 8])
avg_arr = (arr[:-1] + arr[1:]) / 2.0
print(avg_arr)
You can use np.convolve to implement two-element averaging as the convolution of the input vector with the vector [0.5, 0.5]:
>>> a = [1, 2, 3, 4, 5]
>>> np.convolve(a, [0.5, 0.5], "valid")
array([1.5, 2.5, 3.5, 4.5])
np.convolve is a highly optimized routine, and will delegate to different implementations depending on what it estimates will be fastest for the given inputs. For small vectors like [0.5, 0.5] it will likely just compute the result directly (0.5*a[n] + 0.5*[n+1] for each element), but for large inputs it can leverage FFT-based convolution algorithms to compute the result with significantly fewer operations than the direct method.