CODEWITHSUNDEEP
pythonarraysnumpysoftmax
rksh
rksh

Reputation: 4050

Numpy Softmax - The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

I am using a softmax function in getting an output from a neural network and getting the minimum value as the output in calculating the error.

However if the output is all the same assuming [0,0,0] the output of the softmax function is [0.33,0.33,0.33]

So when selecting the minimum from this like, 

output = softmax(np.dot(hs,HO))
tarminout = np.subtract(target,output)
mine = min(tarminout)
mine = 0.5 * np.power(mine,2)
finalError += mine

It gives the following error because there are more than one equal minimum values,

Traceback (most recent call last):
File "ann.py", line 234, in module

learn()
File "ann.py", line 97, in learn

mine = min(tarminout)

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

How can I get pass this when there are more than 1 equal minimum values by selecting just one of them?

Thanks

Upvotes: 2

Views: 5275

Answers (1)

welch
welch

Reputation: 964

The answer's buried in the comments above: your error is likely the result of passing a multi-dimensional ndarray to the standard python min(), which doesn't understand them.

Way #1: call np.min instead of min

Way #2 (not recommended): flatten your array, min(tarminout.ravel())

Way #1 is preferred, use numpy operators on numpy arrays

Upvotes: 2

Related Questions