CODEWITHSUNDEEP
pythonarraysnumpy
LoveMeow
LoveMeow

Reputation: 4038

Computing gradient in python error?

I am trying to follow the course here http://cs231n.github.io/optimization-1/ , in the section Computing the gradient numerically with finite differences, they have provided a code snippet that should computer the gradient given a function and an array. I tried to run this using my own function and numpy array as an input and I get the following error:

ValueError                                Traceback (most recent call last)
<ipython-input-18-31c1f1d6169c> in <module>()
      2     return a
      3 
----> 4 eval_numerical_gradient(f,np.array([1,2,3,4,5]))

<ipython-input-12-d6bea4220895> in eval_numerical_gradient(f, x)
     28     print(x[ix])
     29     # compute the partial derivative
---> 30     grad[ix] = (fxh - fx) / h # the slope
     31     it.iternext() # step to next dimension
     32 

ValueError: setting an array element with a sequence.

I understand the error is because it cannot assign grad[ix] a sequence, I also tried with a column array and got the same error.

Here is the code:

def eval_numerical_gradient(f, x):
  """ 
  a naive implementation of numerical gradient of f at x 
  - f should be a function that takes a single argument
  - x is the point (numpy array) to evaluate the gradient at
  """ 

  fx = f(x) # evaluate function value at original point
  print(x)
  print(fx)
  grad = np.zeros(x.shape)
  h = 0.00001

  # iterate over all indexes in x
  it = np.nditer(x, flags=['multi_index'], op_flags=['readwrite'])
  while not it.finished:
    print(it)
    # evaluate function at x+h
    ix = it.multi_index
    print(ix)
    old_value = x[ix]
    print(old_value)
    x[ix] = old_value + h # increment by h
    print(x)
    fxh = f(x) # evalute f(x + h)
    print(fxh)
    x[ix] = old_value # restore to previous value (very important!)
    print(x[ix])
    # compute the partial derivative
    grad[ix] = (fxh - fx) / h # the slope
    it.iternext() # step to next dimension

  return grad

My question is: is my input of a numpy array (row and column) wrong? Can somebody explain why this is happening?

Sample input :

def f(a):
    return a

eval_numerical_gradient(f,np.array([[1],[2],[3]]))

and

def f(a):
    return a

eval_numerical_gradient(f,np.array([1,2,3]))

Upvotes: 1

Views: 358

Answers (1)

ChoF
ChoF

Reputation: 578

I suggest the following fix for eval_numerical_gradient(f, x):

  • Line #25: replace fxh = f(x) with fxh = f(x[ix])
  • Line #30: replace grad[ix] = (fxh - fx) / h with grad[ix] = (fxh - fx[ix]) / h

And make your input matrix with float number entries, e.g.,

eval_numerical_gradient(f,np.array([[1],[2],[3]], dtype=np.float))

Upvotes: 2

Related Questions