Adaline, (Optimal) Perceptron and Backpropagation Instructions
- To specify a training set:
To train the network:
- Select a category to learn, 1/red or 0/blue.
- Click in the plane to place some points.
- Then change the category and click to place points in the other category.
To compare learning rules, use the same data set each time. When
you change the learning rule, don't forget to click the Init button
Click the Clear button to erase the training set.
- First select a learning rule by clicking in the pop-up menu (near the
top left of the applet).
- Adjust the number of iterations and learning rate parameter in the
fields near the bottom of the applet.
- Click the Init button to initialize the network to a random
set of weights.
- Click the Learn button to train the network. Training
will run for the number of iterations (epochs) you specified; you can continue
training by clicking again on Learn or clicking the Play button. Clicking
the Play button is like clicking the Learn button repeatedly with
a time intervall which you can specify in "delay".
- After training, the color of the input plane indicates the output value
for that point (0.0=blue, 1.0=red).
[Back to the Adaline, Perceptron and
Backprop applet page ]