Skip to content

Latest commit

 

History

History
228 lines (133 loc) · 5.94 KB

numpy_vectorize.md

File metadata and controls

228 lines (133 loc) · 5.94 KB
Raw

Numpy Vectorization : numpy.vectorize() and decoration

https://docs.scipy.org/doc/numpy/reference/generated/numpy.vectorize.html

  • numpy.vectorize() : object의 nested sequence느 numpy array를 input으로 받고, numpy array 또는 numpy array의 tuple을 return하는 vectorized function을 정의함.

Example 1. 비교연산함수

def myfunc(a, b):
  "Return a-b if a > b, otherwise return a + b"
  return a-b if a > b else a+b

vfunc = np.vectorize(myfunc)
vfunc([1,2,3,4], 2)
array([3, 4, 1, 2])
  • docstring : 따로 정의되지 않으면 input function의 docstring을 가져감.
print('1. default docstring (of input) : {}'.format(vfunc.__doc__))

vfunc = np.vectorize(myfunc, doc='Vectorized "myfunc"')
print('2. vectorization function docstring : {}'.format(vfunc.__doc__))
1. default docstring (of input) : Vectorized "myfunc"
2. vectorization function docstring : Vectorized "myfunc"
  • output type: 따로 정의되지 않으면 input의 first argument를 따름.
out = vfunc([1, 2, 3, 4], 2)
print('1. default output type (of input) : {}'.format(type(out[0])))

vfunc = np.vectorize(myfunc, otypes=[float])
out = vfunc([1, 2, 3, 4], 2)
print('2. vectorization function outtype: {}'.format(type(out[0])))
1. default output type (of input) : <class 'numpy.int32'>
2. vectorization function outtype: <class 'numpy.float64'>

Example 2. 다항함수

  • exclude : 특정 argument를 vectorizing에서 제외함.
    polyval의 coefficients처럼 길이가 일정한 array-like arguments를 다룰 때 유용함.
def mypolyval(p, x):
  _p = list(p)
  res = _p.pop()
  while _p:
    p1 = _p.pop()
    res = res*x + p1
  return res

vpolyval = np.vectorize(mypolyval, excluded=['p'])
vpolyval(p=[1,2,3], x=[0, 1, 2])
  
array([ 1,  6, 17])
polynomial의 계수 list인 p를 exclude하지 않으면 아래와 같이 오류가 발생함.
vpolyval = np.vectorize(mypolyval)
vpolyval(p=[1,2,3], x=[0,1])
---------------------------------------------------------------------------

TypeError                                 Traceback (most recent call last)

<ipython-input-23-5466f131c677> in <module>
      1 vpolyval = np.vectorize(mypolyval)
----> 2 vpolyval(p=[1,2,3], x=[0,1])


~\Anaconda3\lib\site-packages\numpy\lib\function_base.py in __call__(self, *args, **kwargs)
   2089             vargs.extend([kwargs[_n] for _n in names])
   2090 
-> 2091         return self._vectorize_call(func=func, args=vargs)
   2092 
   2093     def _get_ufunc_and_otypes(self, func, args):


~\Anaconda3\lib\site-packages\numpy\lib\function_base.py in _vectorize_call(self, func, args)
   2159             res = func()
   2160         else:
-> 2161             ufunc, otypes = self._get_ufunc_and_otypes(func=func, args=args)
   2162 
   2163             # Convert args to object arrays first


~\Anaconda3\lib\site-packages\numpy\lib\function_base.py in _get_ufunc_and_otypes(self, func, args)
   2119 
   2120             inputs = [arg.flat[0] for arg in args]
-> 2121             outputs = func(*inputs)
   2122 
   2123             # Performance note: profiling indicates that -- for simple


~\Anaconda3\lib\site-packages\numpy\lib\function_base.py in func(*vargs)
   2084                     the_args[_i] = vargs[_n]
   2085                 kwargs.update(zip(names, vargs[len(inds):]))
-> 2086                 return self.pyfunc(*the_args, **kwargs)
   2087 
   2088             vargs = [args[_i] for _i in inds]


<ipython-input-22-62002f79cfb3> in mypolyval(p, x)
      1 def mypolyval(p, x):
----> 2   _p = list(p)
      3   res = _p.pop()
      4   while _p:
      5     p1 = _p.pop()


TypeError: 'numpy.int32' object is not iterable

Example 3. Pearson Correlation Coefficient with p-value

  • signature : fixed length의 non-scalar array에 적용되는 vectorizing function을 허용함.
    Pearson Correlation Coefficient와 p-value에 대한 vectorization 예제
from scipy.stats import pearsonr as P

pearsonr = np.vectorize(P, signature='(n),(n)->(),()')
pearsonr([[0, 1, 2, 3]], [[1, 2, 3, 4], [4, 3, 2, 1]])
(array([ 1., -1.]), array([0., 0.]))
vectorization convolution
convolve = np.vectorize(np.convolve, signature='(n),(m)->(k)')
convolve(np.eye(4), [1,2,1])
array([[1., 2., 1., 0., 0., 0.],
       [0., 1., 2., 1., 0., 0.],
       [0., 0., 1., 2., 1., 0.],
       [0., 0., 0., 1., 2., 1.]])

* numpy.vectorize as decoration

@np.vectorize
def myfunc_decvec(a, b):
  "Return a-b if a > b, otherwise return a + b"
  return a-b if a > b else a+b

myfunc_decvec([1,2,3,4], 2)
array([3, 4, 1, 2])
@decoration with options: Error!

Note that np.vectorize isn't really meant as a decorator except for the simplest cases. If you need to specify an explicit otype, use the usual form new_func = np.vectorize(old_func, otypes=...) or use functools.partial to get a decorator.

source : https://stackoverflow.com/questions/14986697/numpy-vectorize-as-a-decorator-with-arguments

@np.vectorize(doc='Vectorized "myfunc"')
def myfunc_decvec(a, b):
  "Return a-b if a > b, otherwise return a + b"
  return a-b if a > b else a+b

print('vectorization function docstring : {}'.format(myfunc_decvec.__doc__))
---------------------------------------------------------------------------

TypeError                                 Traceback (most recent call last)

<ipython-input-36-0cc9eadaea22> in <module>
----> 1 @np.vectorize(doc='Vectorized "myfunc"')
      2 def myfunc_decvec(a, b):
      3   "Return a-b if a > b, otherwise return a + b"
      4   return a-b if a > b else a+b
      5 


TypeError: __init__() missing 1 required positional argument: 'pyfunc'