Network Architectures
1) Feed Forward Networks
2) Recurrent networks
3) Hybrid networks
3.1) Synfire chain
3.2) Liquid State Machine (LSM)
Neural Information Processing
Rate coding
Temporal coding
Learning in SNN
a) Synaptic plasticity (Unsupervised Learning)
Spike Timing Dependent Plasticity (STDP)
b) Supervised Learning
1) SpikeProp
WARNING. SpikeProp can only be used with time to first spike encoding schemes.
|
AND Gate trained using the Izhikevich neuron model and the SpikeProp algorithm.
|
|
OR Gate trained using the Izhikevich neuron model and the SpikeProp algorithm.
|
2) Statistical methods
3) Linear algebra methods
4) Evolutionary methods
WARNING.
- Time consuming computations.
- Only works on single spike patterns.
- Time consuming computations.
- Only works on single spike patterns.
5) Synfire chains
6) Spike based supervised Hebbian learning
References
[1] http://obooij.home.xs4all.nl/nerdspul/spiking_neural_networks/spikeprop/index.html
[2] http://obooij.home.xs4all.nl/nerdspul/spiking_neural_networks/spikeprop/masterthesis/index.html
[3] https://www.youtube.com/watch?v=7QKm2p31bIU
[2] http://obooij.home.xs4all.nl/nerdspul/spiking_neural_networks/spikeprop/masterthesis/index.html
[3] https://www.youtube.com/watch?v=7QKm2p31bIU