Why can’t I raise to a negative power in numpy?

by

In NumPy, the logic used to choose the output dtype of an operation like a_range ** s is based on dtypes, not values. That means that a_range ** -2 has to have the same output dtype as a_range ** 2.

It’s important that something like numpy.array([2]) ** 2 give integer output, so that means that numpy.array([2]) ** -2 has to give integers or nothing. They picked nothing; raising integers to negative integer powers is an error in NumPy.

If you want floating-point output, make floating-point input:

a_range = np.arange(2, n + 1, dtype=float)

or

a_range = np.arange(2, n + 1).astype(float)

There are a few weird aspects of NumPy’s type rules you might not expect from the above description. One is that for operations involving both scalars and arrays, the scalar’s dtype may actually be “demoted” based on its value before the input dtypes are used to choose the result dtype: 

>>> (numpy.array([1], dtype="int8") + numpy.int32(1)).dtype
dtype('int8')
>>> (numpy.array([1], dtype="int8") + numpy.array([1], dtype="int32")).dtype
dtype('int32')

Here, the scalar numpy.int32(1) gets “demoted” to int8, but the arrays don’t get demoted. (It’s actually a bit more complex than just demoting to int8, particularly for signed/unsigned handling; see the implementation for the full details.)

Second, when uint64s are involved, NumPy may suddenly appear to be okay with negative exponents:

>>> numpy.arange(5, dtype="uint64") ** -2
__main__:1: RuntimeWarning: divide by zero encountered in power
array([       inf, 1.        , 0.25      , 0.11111111, 0.0625    ])

This is because NumPy cannot find an integer dtype big enough for both uint64 values and negative values, so it gives up and coerces the inputs to floats. The same can be seen with a positive exponent of signed dtype, as long as you avoid scalar type “demotion”:

>>> numpy.arange(5, dtype="uint64") ** numpy.array([2], dtype="int32")
array([ 0.,  1.,  4.,  9., 16.])

More Related Contents:

  1. Is there a NumPy function to return the first index of something in an array?
  2. How to calculate rolling / moving average using python + NumPy / SciPy?
  3. Saving a Numpy array as an image
  4. Intuition and idea behind reshaping 4D array to 2D array in NumPy
  5. Removing nan values from an array
  6. How to count the occurrence of certain item in an ndarray?
  7. How do I compute derivative using Numpy?
  8. ValueError: could not broadcast input array from shape (224,224,3) into shape (224,224)
  9. How can I release memory after creating matplotlib figures
  10. Sliding window of M-by-N shape numpy.ndarray
  11. numpy array concatenate: “ValueError: all the input arrays must have same number of dimensions”
  12. Precision lost while using read_csv in pandas
  13. reading v 7.3 mat file in python
  14. Numpy: find index of the elements within range
  15. How do I calculate r-squared using Python and Numpy?
  16. How to generate audio from a numpy array?
  17. Fastest save and load options for a numpy array
  18. Pandas: Difference between largest and smallest value within group
  19. Index multiple, non-adjacent ranges in numpy
  20. Change values on matplotlib imshow() graph axis
  21. How to read NumPy 2D array from string?
  22. Python Numpy Data Types Performance
  23. How can I slice each element of a numpy array of strings?
  24. A faster alternative to Pandas `isin` function
  25. Can i set float128 as the standard float-array in numpy
  26. Creating a numpy array of 3D coordinates from three 1D arrays
  27. How to use Python multiprocessing Pool.map to fill numpy array in a for loop
  28. Why is numpy.prod() incorrectly returning negative results, or 0, for my long lists of natural numbers?
  29. Improve performance of converting numpy array to MATLAB double
  30. Faster way to transform group with mean value in Pandas

Leave a Comment