What is a spiking neuron?

Spiking neurons are biologically plausible neurons that use short and sudden increases in voltage to send information, these signals are more commonly known as the action potentials or spikes or pulses. Neurons send out short pulses of electrical energy as signals if they have received enough. When a spike arrives at the axonal (presynaptic) side of the synapse (the gap between two neurons) it is likely that some vesicles fuse with the cell membrane and release their neurotransmitter (endogenous chemicals such as serotonin, dopamine, glutamate, etc that allow neurons to communicate with each other) content into the extra-cellular fluid that fills the synaptic gap. Such potential can either be positive called excitation or negative called inhibition.

Neuromorphic Computing

• Neuromorphic computing is the type of computation in which the elements of hardware are modeled after systems in the human brain and the nervous system.
• A Neuromorphic computer should have the following characteristics:
  • Collocated processing and memory: The brain-inspired neuromorphic computer chips process and store data together on each individual neuron instead of having separate areas for each.
  • Massively parallel: Each neuron operates on different functions simultaneously. This type of parallel functioning mimics the stochastic noise, which is the seemingly random firing of neurons in the brain.
  • Inherently scalable: To run larger networks, often more neuromorphic chips are added.
  • Highly adaptable: Like humans, neuromorphic computers are designed to be flexible to changing stimuli. In SNNs, each synapse is assigned a voltage output and is adjusted based on the task given.
  • Fault tolerance: Like the human brain, information is held in multiple places, the failure of one component does not prevent the computer from functioning.

The “brain” Theory

The brain is the most complex organ in the human body. It is made up of billions of specialized cells, including neurons, the cells that transmit and receive information and signals. Glial cells perform multiple support functions as well as contribute to signaling and other cells that make up the blood vessels and other structural elements.

Neuron structure:
Each neuron is composed of several different parts. The nucleus, which is located inside the cell body called the soma, contains the cell’s genetic material and is responsible for controlling the cell. Dendrites are branch-like structures that project out of neurons and taper towards the end. The axon is a thin cylindrical projection from the cell body that resembles a tail. The axon also splits into multiple branches, each of which has an ending called an axon terminal, where the transfer of information from one cell to another takes place.

Ion channels:
The permeability of cell membranes to certain ions is achieved via ion channels. Ion channels are special types of proteins embedded in the membranes of the cell. Common ions involved in such processes within the nervous system are Sodium (Na+), Potassium (K+), Calcium (Ca2+), and Chloride (Cl-). The ion channels that drive the resting potential of neurons are usually open all the time. It is hence appropriate to call them leakage channels. There are many types of ion channels, covering them would be out of the scope of this article. Some types are:

1. Leakage channel
2. Voltage-gated ion channel
3. Ion pump
4. Ionotropic
5. Metabotropic

Associative memory and Hebbian learning

Neural plasticity refers to the capacity of the neuron to modify itself, functionally and structurally, in response to the stimuli. Structural plasticity is the mechanism describing the generation of new connections thereby redefining the topology of the network. Whereas functional plasticity is the mechanism of changing strength values (weights) of existing connections.
Hebbian learning:
When an axon of cell A is near enough to excite cell B or repeatedly or persistently takes part in firing it, some growth or metabolic change takes place in both cells such that cell A’s efficiency, as one of the cells firing B is increased.
Hebb had no means of observing synapses directly, which makes this hypothesis a wild guess by our standards. Activity-dependent plasticity that depends on pre- and postsynaptic activity is often called Hebbian plasticity of Hebbian learning.
Associations:
Synaptic plasticity is the necessary ingredient behind forming associations.
“Increase the strength of synapses by a value (w = 0.1) if a presynaptic firing is paired with a postsynaptic firing”. This learning rule modifies the synaptic weights so that after one simultaneous presentation of two or more stimuli, the respective cue will find it. Although the synaptic strengths of input channels change with the learning rule, we call the initial stimulus the unconditioned stimulus because the response of the neuron to this stimulus was maintained. For the second input the response of neuron changes during learning, and so we call the second stimulus the conditioned one.

Spiking Neuron model

A spiking neuron has a similar structure to that of a perceptron (artificial neuron) but shows different behavior. Different models differ not only on which biological characteristics of real neurons they can reproduce but also on their computational complexity.

There are many types of spiking neuron model such as the hodgkin-huxley model, Izhikevich model, Integrate-and-Fire model, and so on.. I have written a descriptive article describing each model in detail on medium. Do check it out!
I have also made a github repository, implementing each algorithm and models. Explaining the processes bit by bit. Stay tuned for exciting updates!!

Bottom-up!
~Anirudh.