Problem 58882. Neural Nets: Activation functions

Created by Richard Zapor

Solve Later

Return values of selected Activation function type for value,vector, and matrices.
y=Activation(x,id); where id is 1:4 for ReLU, sigmoid, hyperbolic_tan, Softmax
ReLU: Rectified Linear Unit, clips negatives  max(0,x)   Trains faster than sigmoid
Sigmoid: Exponential normalization [0:1]      
HyperTan: Normalization[-1:1]   tanh(x)
Softmax: Normalizes output sum to 1, individual values [0:1]      Used on Output node
Working though a series of Neural Net challenges from Perceptron, Hidden Layers, Back Propogation, ..., to the Convolutional Neural Net/Training for Handwritten Digits from Mnist. 
Might take a day or two  to completely cover Neural Nets in a Matlab centric fashion. 
Essentially Out=Softmax(ReLU(X*W)*WP)

Solve

Solution Stats

66 Solutions
13 Solvers

Last Solution submitted on Nov 12, 2025

Last 200 Solutions

Problem Comments

1 Comment

1 Comment

Richard Zapor on 21 Aug 2023

Multi-Case Softmax should be y=exp(x)./sum(exp(x),2)

Solution Comments

Show comments

Problem Recent Solvers13

Problem Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Problem 58882. Neural Nets: Activation functions

Solution Stats

Last 200 Solutions

Problem Comments

Solution Comments

Problem Recent Solvers13

Suggested Problems

More from this Author308

Problem Tags

Community Treasure Hunt

Problem 58882. Neural Nets: Activation functions

Solution Stats

Last 200 Solutions

Problem Comments

Solution Comments

Problem Recent Solvers13

Suggested Problems

More from this Author308

Problem Tags

Community Treasure Hunt

Players