Adding layers

This dataset is more difficult than the last one. A single line can no longer separate the red dots from the blue ring. To solve more complex problems, the neural network needs to be larger.

You can now add another layer to the network. Layers between the input and output layers are called hidden layers.

Adding neurons

You can also add new neurons to hidden layers. Try it out! Click on a neuron to change its weights and bias. More neurons allow the network to create more complex decision boundaries.

User Controls

You can now control and train the network.

The overview shows the network structure, completed epochs, and current accuracy. One epoch means the network has trained once on the entire training dataset.

The green button starts training. The red button resets all weights and biases to random values.

The training settings let you choose the learning rate and number of epochs. The learning rate controls how much the network changes during each training step.

Helpful notifications

If the network gets stuck and stops improving, helpful tips may appear in the top-right corner.
Read them carefully, as they may help you fix the problem.
Click on a message to close it.