Spiking Neurons and Networks, and Spike Timing Dependent Plasticity
- Spiking Neurons and Networks
- Brains work with neurons that collect activation and
if they get enough, they fire.
- They're a lot like binary output perceptrons, but they work in
continuous time.
- There are a lot of different models of neurons, but we work with
point models.
- The simulations here use relatively standard leaky integrate and
fire (LIF) neurons, and
- relatively novel LIF neurons with dynamic thresholds.
- A spiking network is just a bunch of neurons connected with
synapses.
- We use uni-directional synapses and these take time to spread
the activation.
- STDP
- In the brain long term plasticity is Hebbian. (If neuron A tends
to cause neuron B to fire the weight tends to go up.)
- Spike Timing Dependent Plasticity (STDP) has some pretty strong
biological evidence.
- We use a relatively novel form of STDP, though see the
After Paper page.
- That's the only type of learning in the system.
- We use PyNN and the Nest simulator.
