Learning to Classify Students
- A simple two level all to all topology is used for this task.
- Each input feature value has a neuron associated with it as does
each category.
- Each input neuron is connected to each output neuron with an STPD
synapse.
- For training, for each item, the input neurons are fired, followed by
the output category.
- Weights grow to the point where the input itself can cause the
output category neuron to fire.
- Testing presents new input with no information about the output
neurons.
- The category is the first ouput neuron to fire.
- Different encoding methods led to different results with the
best producing undone correct categorisations.
- With longer learning windows, and encoding with one hot encoding,
up sampling and no scaling the system performed best getting
72.4%.
