Make your own free website on Tripod.com

Background

Program objective: To write a program to classify two types of 1-D curve (∩-shape or U-shape) by using a perceptron, i.e., only one neuron.

Structure of neuron:

 

For perceptron, i.e., only one neuron, it can classify two class, '∩' or 'U' curves.

The formulue we used are:

1. v = Σ x(i)w(i)
2. y = {-1 if v<0, 0 if v=0, 1 if v>0 }
3. Δw(i)=dx(i) for d ≠ y as learning rate is 1
4. new w(i) = old w(i) + Δw(i)

where i is from 1 to 6.

 

The 5 default training sets used in this program:

set 1 : {0.76, 0.65, 0.50, 0.23, 0.37, 0.48}
set 2 : {0.18, 0.30, 0.40, 0.52, 0.42, 0.25}
set 3 : {0.44, 0.49, 0.58, 0.40, 0.48, 0.12}
set 4 : {0.29, 0.18, 0.11, 0.18, 0.14, 0.71}
set 5 : {0.37, 0.42, 0.53, 0.54, 0.46, 0.23}