Neuromorphic Computing Journey (Part 3)
Hello everyone to the 3rd part of this series.
Brief recap. Till now we have discussed history of Neuromorphic Computing and then the projects related to this field, including SpiNNaker: Multi-core System-on-chip. In this part we will discuss the software aspects of SpiNNaker and will jump into what are the Spiking Neural Networks and why they are important.
Before getting into the programming side of the SpiNNaker. First have a look into what happens when this million core machine is booted. This flow of procedures are as follows:
To work with SpiNNaker Machine, you don’t need to know the whole process. I added this flow diagram just to give you an overview. To execute code on local SpiNNaker board (SpiNN-3, SpiNN-5), you have to follow these steps:
- Boot the SpiNNaker machine
- Load executables onto the SpiNNaker machine
- Load data objects onto SpiNNaker
- Check when the executing code finished
- Extract data from the SpiNNaker machine
Spiking Neural Network
Our brain has million and million of neurons connected to each other which help us in processing the information, decision making etc. As you might know the general idea of Artificial Neural Network came from biology. But in reality artificial neural network do not mimic the behaviour of the biological ones.
This is how neuro transmission in our brain looks like:
And this is how Artificial Neural Network (ANN) looks like:
How Spiking Neural Network is different then ANN. Though both are inspired from how human brain works but SNN actually try to mimic the behaviour of brain. Our brain works on spikes. Whatever you feels or think, there is a pattern of spikes generated in you brains which is then processed by neurons in brain.
In architecture of SNN is also similar. You will get better understanding with following image:
This is just glimpse of SNN. In next blog i’ll go deep into the concept of Spiking Neural Network.
Stay tuned.
