Compute softmax activation function using python

Question

I was trying to write a method to compute the SoftMax activation function that takes either a matrix or an array as input and apply the softmax function to each rows.

Here is what I tried:

import numpy as np
def softmaxSingle(x):
    e_x = np.exp(x - np.max(x))
    return e_x / e_x.sum()    

def softmax( x):
    if np.shape(x)[0]>1:
        result=[[]]*np.shape(x)[0]
        for i in range(len(result)):
            result[i]=list(softmaxSingle(x[i]))
        return list(result)
    e_x = np.exp(x - np.max(x))
    return e_x / e_x.sum()

When I tried SoftMax(x) where x is a matrix, It runs(although I don't know if it produces correct answer). When x is just a list, it doesn't work

pissall · Accepted Answer · 2019-11-17 18:49:28Z

1

You can simply make a list to np.array conversion:

import numpy as np

def softmax(x):
    """Compute softmax values for each sets of scores in x."""
    if isinstance(x, list):
        x = np.array(x)
    e_x = np.exp(x - np.max(x))
    return e_x / e_x.sum()

answered Nov 17, 2019 at 18:49

pissall

7,4442 gold badges29 silver badges47 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

user42493 Over a year ago

But 'x-np.max(x)' method doesn't substract each row by the max of each row, it substract each row by the max of all entries, Is that ok? Try this for example: a=[[1,2,3],[4,5,6]] print(a-np.max(a)) and the output of that is : [[-5 -4 -3] [-2 -1 0]]

pissall Over a year ago

@user42493 Can you give me sample input? And yes, it needs to subtract every element by the max of that matrix, don't you think? If row-wise is a requirement then you can still work it that way

user42493 Over a year ago

Ok, what about e_x.sum() ? that gives the sum of all entries in the matrix, but we need to divide each row of e_x by the sum of each row e_x?

pissall Over a year ago

From the docs: Axis or axes along which a sum is performed. The default, axis=None, will sum all of the elements of the input array. If axis is negative it counts from the last to the first axis.

Collectives™ on Stack Overflow

Compute softmax activation function using python

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related