Skip to content

Update 1.3 Shallow neural networks.md #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
dani-capellan wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Update 1.3 Shallow neural networks.md #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions 1.3 Shallow neural networks.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,20 +87,20 @@ For example, if X has 3 training examples, each exmaple with 2 values.

- $a = tanh(z) = \frac{e^z-e^{-z}}{e^z+e^{-z}}$
- Range: a [-1,1]
- Almost always work better than sigmoid because the value between -1 and 1, the activation is close to have mean 0 (The effect is similar to centering yhe data)
- Almost always work better than sigmoid because the value between -1 and 1, the activation is close to have mean 0 (The effect is similar to centering the data)
- If z is very lare or z is very small, the slope of the gradient is almost 0, so this can slow down gradient descent.

- ReLu

- $a = max(0,z)$
- Derivative is almost 0 when z is negative and 1 when z is positive
- Due to the derivative property, it can be faster than tank
- Due to the derivative property, it can be faster than tanh

- Leacy ReLu
- Leaky ReLu

Rules of thumb:

- If output is 0,1 value (binary classification) ->sigmoid
- If output is 0,1 value (binary classification) -> sigmoid
- If dont know which to use: -> ReLu

## Why non linear activation function?
Expand Down Expand Up @@ -211,4 +211,4 @@ The general methodology to build a Neural Network is to:
predictions = (A2>0.5)
```