error in numpy:TypeError: only length-1 arrays can be converted to Python scalars

Question

I write a function sigmoid:

def sigmoid(inX):
    return 1.0 / (1 + exp(-inX))

and when the inX is a matrix of 100x1 size, the error arises:

/home/abyss/python/machine/logistic/logReg.py in sigmoid(inX)
     12 
     13 def sigmoid(inX):
---> 14     r = 1.0 / (1 + exp(-inX))
     15     return r
     16 

TypeError: only length-1 arrays can be converted to Python scalars

but I can directly use this expression in command line:

In [59]: r = 1.0 / (1 + exp(-h))

In [60]: shape(r)
Out[60]: (100, 1)

I am totally confused, how did this happen?

Answer 1

When you are taking the exponent of an array or anything else you don't want to use Python's math library, use NumPy library instead:

import numpy as np
def sigmoid(inX):
    r = 1.0/(1+ np.exp(-inX))
    return r