Neural Networks Fundamentals Explained

Neural Networks Fundamentals Explained

Introduction

We all know that the human brain is a complex structure consisting of millions of neurons which perform the function of thinking, reasoning etc which we call intelligence. But do you know that similar intelligence can be imitated in machines albeit to a much lower degree of sophistication. Such machines which are based along the lines of the human brain are said to possess artificial intelligence. We will learn about the fundamentals of neural networks in this subsequent sections.

Neural Networks Fundamentals

So basically a machine containing artificial intelligence is build based on neural networks which try to emulate the neural network in the human body, hence the nomenclature. Although it is too early to discuss about the exact construction of neural networks, they are basically constructed out of simple processing units, which are combined in massive numbers to form huge networks.

Hence in a way neural networks are massive parallel processing units which can perform tasks which were not possible with the conventional computing techniques. The conventional computers were mainly based on imitating the human thought process, but neural networks imitate the manner in which it is achieved in the human brain.

Hence the machines consisting of neural networks based artificial intelligence do not need to be precisely programmed for each and every detail. They are quite similar to humans (I mean only to a very limited extent at the current level of development of this technology) wherein you only teach a person how to do a task in a general manner and then the person uses that previous knowledge, reasoning, experience to find innovative solutions to new problems.

Of course I must say that even we think that technology has advanced to such great heights, still the human brain is much more superior to even the latest artificial intelligence machines available anywhere on the planet today.

The only factor where the electronic devices take lead is the speed of operation which is around six orders of magnitude higher than human neurons. As for the rest, the artificial intelligence machines are still nowhere near even a less developed human brain which is par-excellence in non-linear parallel information processing, pattern recognition, reasoning and so forth. Given below are a few points which give an indication of the areas in which neural networks are inferior to the biological neural systems.

Artificial neural networks are not fault tolerant in a way that once information gets corrupted in the memory it is not possible to retrieve the same, while this is not so in the human brain mainly because of the distributed nature of storage.

There is a central control unit for the entire network whilst it is not so for the human brain. Of course this may not be always a disadvantage but also means more control over the processes.

The size and complexity of an artificial neural network cannot resemble the human neural network which has nearly 1011 neurons approximately.

And although this is not a biology related article, still I would like that you take a look at the rough sketch of a neuron. This will help you to compare it to the relevant electrical circuit later on.

As you can see in the sketch, basically a neuron consists of a

  1. Cell body called soma
  2. Input channels called dendrites
  3. Output channel called axon

These neurons are connected to several other neurons and form the network in this manner.

Of course the fact that artificial intelligence created using neural networks does not exactly imitate the brain does not mean that it is useless. There are lots of uses for artificial intelligence and we will study about them in later articles. You can read an article on fuzzy logic here.