Can you imagine replicating the human brain? Think its impossible?..Well...think again. With artificial neural networks its POSSIBLE!! See, the human brain is truly special. Haven't you wondered how come you can see your friend's photo and remember "oh yeah..he's my buddy"? Well we do know that humans have a capacity to learn and remember. These two highly developed qualities are what set us apart from the rest of the creatures gracing mother Earth. We are already familiar with various programming languages like C,C++,Java and a whole lot of other languages that help us solve problems. I can perform complex calculations, logical analogies and a truckload of other little miracles using these superb techniques. But is it possible to create a replica of the brain? Something that can actually not only analyse, but learn and remember at the same time? Something that we can train to do what we want? The answer was 'no' until artificial neural networks, more popularly known as ANNs came into being!

       You will be as astounded as I was to learn that ANN research began in 1943, years before traditional computing actually began!! It was left in the dark may be because majority of researchers felt that it was a useless plan. "How could the human brain be mimicked?"they thought(may be...I am just making guesses here). In an age where even going to space was being accepted, the replication of the human brain seemed a legend that would never take fruit. All the brilliant minds shifted back to computing..a field that has developed and progressed..leaving behind the beautiful ANN concept decades behind and just still emerging from its cocoon. Well don't be too depressed because ANNs came back with a bang and how!! Mimicking the human brain is no longer a far fetched dream, but a reality that you can grasp if you are willing enough to reach out to it. And how exactly is this done? Simple really. Replicate the human neurons with sophisticated technology. Lets not delve into the hyper complicated technological know-hows of how all this happens. Lets talk in simple layman terms.

       Now, anyone with an high school degree ought to know that the human brain consists of special cells called neurons with really long tails(axon). These axons touch the other neurons and information is transmitted as nerve impulses which are just plainly chemical reactions. Similarly in the ANN model we have primary neurons that are linked up with other neurons and these in turn with others and this process goes on. Now each connection is called a weight and by manipulating this we can determine what the network is supposed to do. Okay, let us see a simple example and mind you I cooked up this example by myself from what I understood by extensive reading. Please do not take this as the written word. So lets discuss addition of two numbers in layman language. The two numbers are given as the inputs to our first layer of neurons. Next we have to train the neural network what it has to do. This is done by teaching it what to do with examples. Say we gave example number 1 and 2. The result should be 3. But suppose our ANN generated 2.89. Then we vary the weights until the output becomes 3. Next we give another example pair 2 and 3. The output should be 5, but say we get 4.93, then we again vary the weights till we get 5. Now this is done until the ANN is able to do its own calculation. Now give it any number and VOILA!! Our ANN knows what to do and does it correctly because of its memory! Isn't that just splendid. Clap people! We just created a mini brain!

       Now many will be thinking why should we train our ANN to do math. We have calculators for that! Well, you are right! But then can our calculators do detect and recognise faces or predict the stocks for that matter? No right? Hehe..that is where our ANN comes into play. With memory and an ability to learn, its on a whole different platform from anything the world has witnessed till date! Remember an old laptop add, where the laptop recognises Saif Ali Khan though he is atrociously malformed(for lack of a better word)? Well, you too can rig up your pc to do the same and that also without having to purchase hyper costly laptops. Just imagine how cool it would be if your younger sibling tries to take a sneak peek at your pc but they get automatically rejected? Ah, bliss I tell you..sheer bliss!!

       Now if you have become hyper excited and want to try your hand at this, the simplest way will be to learn matlab which is a simple software. It is not a freeware and that is the only bad point that I have found about it. You can actually generate a ANN program using matlab. And 'NO' you don't have to go to expensive tutors, you can just learn from the help section which is just too AWESOME!! Well, I am doing that right now and hopefully I can get an awesumly fabulous face recognition thingamagig made in the near future preferably in two months(I must be dreaming). Anyways for anyone looking for more information Wikipaedia and the neural network book by Simon Haykin will be a huge help! I have imparted whatever knowledge I had. And trust me its not much! Neural networks is a huge sea filled with details, facts and other mysteries! While you get busy demystifying neural networks let me depart with a "Thats all folks!".


Like it on Facebook, Tweet it or share this article on other bookmarking websites.

No comments