Simple Neuron

See a neuron attempting to learn AND, OR or XOR. XOR Sep has the extra input. Two non logic data sets are also available.

By default this is a neuron with linear activation, where sigmoidal can be selected instead, but in addition the neuron can be configured as a McCulloch-Pitts cell (MCP cell), ie where the weighted sum is thresholded. You can also set the learning rate and momentum.

Test Net applies the next line from the training set. The inputs, target, output and Delta values are shown.

Teach Net applies the next item of data and adjusts the weights. When all items in the set are presented, an Epoch, the sum of the squares of the errors (SSE) is shown.

Auto Teach repeatedly applies the data, adjusting the weights. It will stop automatically if the specified number of epochs is exceeded or the SSE is less than the specified value. When you press Stop Teach the learning stops. If you then press Auto Teach again, learning will resume.

When teaching has stopped, the variation of SSE vs Epoch is plotted. Also shown is the Feature Space, where the line associated with the current weights is also shown.

Teaching will also stop when the epochs taken exceeds the maximum specified.

Reset reinitialises the weights