Networks types

• Feedforward versus recurrent

  • Feedforward: No loops (other than error signal)
  • Recurrent: Use feedback (positive or negative)

• Continuous versus spiking

  • Continuous networks model mean spike rate
    • Assume spikes can be integrated
  • Same as temporal-code versus rate-code arguments from biology

• Supervised versus unsupervised learning

  • Supervised networks use a teacher
    • There is a training dataset
  • Unsupervised networks cluster input data

Previous slide

Next slide

Back to first slide

View graphic version