# Alternativa to softmax?

**URL:** https://discourse.julialang.org/t/alternativa-to-softmax/39671
**Category:** Machine Learning
**Tags:** question
**Created:** [May 18, 2020, 2:11am UTC](https://discourse.julialang.org/t/alternativa-to-softmax/39671 "2020-05-18T02:11:53Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![BMval](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/bmval/32/7647_2.png) [@BMval](https://discourse.julialang.org/u/BMval)
#### Post date: [May 18, 2020, 2:11am UTC](https://discourse.julialang.org/t/alternativa-to-softmax/39671/1 "2020-05-18T02:11:53Z")

</div>

alternativa to softmax?

Would you suggest it?

I don’t like softmax since softmax(const \* x) != softmax (x) .  
For instance softmax([0,0.2,0.05]) = [0.3,0.37,0.32], so I can’t interpretare it as probability at all, bacuse value zero gives 0.3.

Thank you in advance  
P.S. I use prob(x\_i) = x\_i/sum(x\_i) right now, it better fit my need.

---

<div class="post-metadata">

### Author: ![xiaodai](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/xiaodai/32/15937_2.png) [@xiaodai](https://discourse.julialang.org/u/xiaodai)
#### Post date: [May 18, 2020, 2:55am UTC](https://discourse.julialang.org/t/alternativa-to-softmax/39671/2 "2020-05-18T02:55:51Z")

</div>

yeah well. You are saying `softmax` doesn’t respect the scale.

What r u using it for? For loss function in nn your `prob` will have issues if `x_i` is negative which `softmax` handles well.

Anyway, it’s not clear what u r using `softmax` or `prob` for.

Overall the question is very vague.