Neural Networks: Chapter 1 - Biological Neuron Vs Artificial Neuron

Neural Networks Jun 7, 2021

Let's start with an unconventional approach to understand Neural Network. As a part of a quote from the movie - Lucy (2014)

Brains in formation of only a few milligrams - it's not possible to determine any signs of intelligence yet, it acts more as a reflex.
One neuron you are alive,
Two neurons - you are moving

And with movement interesting things begin to happen

That being said our Nervous system is a collection of Billions of Neurons, now imagine the possibilities of what we can achieve and we already have. The next obvious question in our mind is -

What is a Neuron?

Simply put -

A neuron is the most fundamental unit of any living organism's Nervous System.

A biological neuron primarily consists of 4 components -

  1. Dendrites - The Receiver
  2. Soma - The processor
  3. Axon - The Catalyst
  4. Synapse - The Transmitter

In an ideal world we humans just like any other machine, transfer data in BINARY across neurons. What's even more note worthy is that, all the Binary information is passed as Electrical impulses - Chemical Reactions generating IONS. This complex web of neurons is what is responsible for the functionalities of any Living Organism.

Now that we think about it, we humans are not so different from machines or may be we built machines to imitate humans. I think it's fair to assume that the later holds true.

So then the next obvious question is -

What is an Artificial Neuron?

An Artificial Neuron is the most fundamental unit of an Artificial Neural Network

An entity that was primarily built to process information that has certain performance characteristics in common with Biological Neurons. Well a broader question to ask would be Can we replicate Human Conscience? But that's more of science-fiction than reality for now. So the real question is Can we replicate simple Human Tasks using Artificial Intelligence?

Well the answer is obviously Yes We Can, or else we wouldn't have this article. But then the next obvious thought is HOW?

An Artificial Neural Network comprises of what we call as Nodes and a complex network of these nodes enable us to perform complex tasks. In addition to that every node has something called as activity state, which is also defined as Activation Function of the node.

Now if we go into further details, a node comprises of -

  1. Axon - The Catalyst - Weights and Biases
  2. Soma - The processor - Activation functions and Matrix Multiplications

Just like in a biological neuron, the chemicals react to activate the neuron and transmit signals, in the same manner the nodes perform complex mathematical operations to trigger the node's activation function and transmit information.

Intuitive Understanding

Think of your node as a Satellite and your Neural Network as a complex web of Satellites. Every time you want a task to be performed, the Satellites need to communicate within themselves via their directed communication links, which we have referred to as Weights and Biases or The Catalyst.

This network is responsible for transmitting information from someone sitting in US to all the way over to someone in Japan. Now even if a single Satellite fails to get triggered, there is a high likeliness that some of the significant information may be lost. This information is what we call as Features.

When a Satellite is triggered it transmits information to multiple Satellites and whichever Satellite is in an Active State transmits it further. This chain of communication makes sure that the information is transformed and transmitted in the expected format to the receiver.

In the same manner your Artificial Neural Network passes information from one node to another and transforms and analyses the information and finally portrays it out to the human cognitive sense in the expected manner.

In reality there is more similarity between Biological Neurons and Artificial Neurons than you might think to be True -

  1. The processing element receives many signals
  2. Signals may be modified by a weight at receiving synapse
  3. Under appropriate circumstances the neuron transmits a single output to multiple neurons
  4. Information processing is local - although hormones play a significant role, which are also coined as Biases in our system
  5. Memory is distributed
  6. Synapse's strength may be modified by experience

Conclusion

One might think that attaining such a feat might be next to impossible for someone with less experience in this field. But not to worry, in our future posts we will be diving deeper and simplifying the concepts of Neural Networks and answering some of the mind boggling questions to bring this distant Dream Come True.

STAY TUNED for more content on Neural Networks. 😁

Tags

Vaibhav Satpathy

AI Enthusiast and Explorer

Great! You've successfully subscribed.
Great! Next, complete checkout for full access.
Welcome back! You've successfully signed in.
Success! Your account is fully activated, you now have access to all content.