Where can I find some good theoretical descriptions about Spiking Neural Networks ? I'm unable to gather resources regarding this kind of network what people call the third generation ANNs (don't even know why).
Asked
Active
Viewed 355 times
1
-
I would personally recommend you to read up on CTRNN. They are capable of simulating all of the spiking patterns shown in the brain. I took part in a project at the university which applied them to play the 8bit game "Space Invaders" – jorgenkg Jan 02 '16 at 22:02
2 Answers
3
Actually, it's an active research area with many new efficient software and hardware implementations. On the software side, look at NEST and BRIAN which are easy-to-use, even though the standard in terms of biological relevancy is NEURON.
All those lack a learning paradigm, which would allow you to tailor a network to solve some computational problem. This is still an open problem. One research direction which already can be used for application is that of (so-called) "echo state network" or "liquid state network". This is a learning algorithm which can be used with spiking neuron simulations, with some effort.
Uri Cohen
- 3,488
- 1
- 29
- 46
1
Spiking neural networks are closer models to real neurons in brain. They have the ability of plasticity to change their characteristics over time. Therefore, they tend to mimic the synaptic plasticity in real neurons in brain and change their activations, architecture and outputs of neurons over time. As a small-scale example of these models you can take a look at the neural networks models whose architectures are learned using evolutionary algorithms (Genetic Algorithms for instance). However, due to lack of enough computational power people have not been able to deploy these models in large-scale applications. There is a very small research going on for these models too.
Sources
