Going to write this article, I remembered about my 5th grader, at that time I was a fine expert in simple math and interest calculation. So being fascinated, the math teacher of the class had said, “Your head is not just head, but it is totally computer.” Though at that time except for “Computer” word, I didn’t know anything more than this, only I knew, all wonderful things are done with it. But that time though I didn’t know, now I know- in fact, the human brain is the most amazing thing. A computer has no intelligence like a man, this machine maintains only order like a firming ox.
But, if the computer is wanted to make intelligent and clever, then how that will be possible? Scientists have run behind this question, and have discovered artificial intelligence, machine learning, and deep learning technology. After many decades of worship, there was a complicated matter to teach computer like men. But with help of neural network, a computer can work same as human brain by using million artificial brain cell like men. The neural network is also well-known as the artificial neural network – which is a part of deep learning. So what is this actually and how does it work? Let’s know about this….
Man’s brain vs. Computer
Brain vs. Computer:
How many animals will come in your imagination, like- mammals, birds, reptile, fish, amphibian etc. everybody has a brain. But, a human brain is completely different than all. Though it is not so big in size, but it helps to talk, to think, to solve problems and it is an important and amazing organ of human body. There is 100 billion atomic cell in a typical brain (but there may be up to 50 billion to 500 billion) – which is called Neuron.
Each neuron has body cells and there are many dendrites (It carries cell’s input until body cell) and one axon (It releases output of information from the cell). Neurons are very small in size, almost 100 neurons can store in each millimeter. There is a location of neuron around 10% of the total cell of a brain, left cells are called Glial Cells or Neuroglia, these supports and protect neuron and transmit power in the neuron.
On the other hand, there is set a very small switch on the computer instead of the human brain, which is called transistor. There are set transistor above 2 billion in the latest microprocessor of the present technology. There are set 50 million transistors in normal microprocessors and all these things are set on one circuit together, which is 25-millimeter square.
So this is human brain and the main structure of a computer. A computer is a metallic box and it works on binary number and brain is a live thing which works on feeling and memory – this is not the only one difference between computer and brain. The real difference is- computer and brain work in totally different process. Transistors are set easily in the computer processor, each transistor is connected with 2/3 transistors, it is called Logic Gates. But neurons are connected very complicatedly with each other parallel. Each neuron is connected with 10,000 neurons. So several hundred million transistors of a computer are connected parallel easily and there is 10-100 times more cell in a brain which are connected complicatedly with each other – for their different structure, computer and brain think and work differently. A computer is designed for preserving large amounts of data (which is meaningless to computer) and how to process data by a computer – programs are installed in the computer for saying that.
On the other hand, a brain gets late to learn anything, it has to practice for keeping in mind and can learn several things in several ways, can make itself the process of new learning – in this way creativity is found in man’s work. A human brain can remember any pattern, can match after, and can see known things in different looks.
So what would happen, if a computer would work like the human brain? And neural network comes here.
What is Neural Network?
If you wanted to make a computer like human brain then it will be required to build it exactly like brain according to the idea. In another word, brain cell has to set in large number in it which will be related to each other – through which pattern is identified by the computer, anything is learned and can take the decision on any matter like a man. The funniest thing of the neural network will be, the computer will not need to make any program for working anything in it, and it will learn work itself like us.
But to speak the truth, it is not possible to make design any computer exactly like a human brain, so not hardware, rather by using software normal transistor setting computer-is tried to behave like billion brain cell setting computer. Which computing we do in present and we work by using which program, its output keeps value only to us, not to a computer. A computer never knows, what you have done with it. Any task that if you give a computer to complete- the computer will calculate that, will change into binary, processing numbers it will decorate into several patterns and you can see the output. Suppose about data storage, in it, you can preserve music, video, photos and more anything. Each data has the different meaning to us, but music, photo, videos are nothing without one and zero to your phone or any computing device.
So mainly neural network is such a computing management which will understand everything like men, will be able to learn from it and will be able to correct any mistake slowly. Not by seeing any dog photo just, the computer will see it as dog photo like men – and it is the neural network. And as the neural network is conducted by software, it is conducted like real brain so this total process is called artificial neural networks or ANNS. Though it works like a human brain but it is not a brain.
How does neural network work?
See its working process or it is complicated to understand happened matters inside it, such many terms will come in it with which it will require discussing differently. So not going to the technical side so much, we will try to understand total matter through easy language and enough example.
There are many artificial neurons from thousand to lakh in an ideal network, these are called Unit. These units are decorated gradually with each other and each is connected to each other exactly like any network. There are some units in it which accept several types of information, like we see anything, try to identify colour, understand drawing-these units are called input unit. These units help to identify, know or process neural network. There is output unit in another side of the network – which publishes it’s outside that exactly what knowledge computer earned from a real thing. But it is not the end, there is a hidden unit in the middle phase between this input unit and output unit, which is structured by most of the artificial neuron of the network. So it was seen that there are three types of unit in it, its group is called layer. Units of each layer are related to each unit and by processing information from each other, after then it gives last result or output. The connection of each other of each unit is drawn by one number – it is called weight. If one unit gives permission with another unit’s information then weight becomes positive and if permission were rejected then weight becomes negative.
Now get back to another point that is – how does it work about that, any data has flowed in the neural network in two ways. When a computer takes the lesson about anything then one type of data has flowed and after taking the lesson, when a computer starts in action, that time one type of data has flowed. When that computer starts to take any lesson then it needs feedback again and again to correct its learning that has it learned right or wrong? Like, if you write wrong spelling in front of your teacher then he corrects your mistake and it requires feedback always for learning something for us. Suppose you are shooting the balloon with an airgun, you shoot at the side of the balloon by aiming the gun, but one bullet went over that balloon. Now second time when you will shoot then certainly you will keep in your mind about previous position and target. Now you certainly will shoot the gun a little bit get down, because your brain knows, he failed last time to shoot up. In this way, you will able to make a good aim by trying and correcting the mistake.
Neural network also learns education exactly like that way. In every time it takes information from the input units. Suppose computer is shown a cat’s photo from the input. It preserves this information in another word, it means to say that how it looks like, what type of size it is, where are eyes, where is mouth etc. Now, each cat is not of the same type, so for teaching network, it needs to show many photos of the cat. Now, if a computer is said by showing another photo, is this cat or dog. But first input unit will accept its information and that will send to the hidden unit. The hidden unit has much information from the previous cat’s photo, so it will start input to analyze. Suppose one unit will observe its nose shape, size and colour etc. if that matches with that cat then that will give positive weight, another unit will observe eyes, another will may tail. So this will give output by depending on this hidden unit’s processing and the result of weight’s positive and negative. If output comes wrong then computer system has to inform that it has given the wrong result, in the next computing system will not do that mistake. Like we take the lesson from the mistake.
Now, suppose you have taught neural network by showing some chair and table’s picture. But now you upheld totally some new model’s chair table picture, which had not been seen before. Now computing system will take the concept from the previously shown chair tables (like men do) and it will divide new chair tables into category differently. But that is also the important matter that which type of education you have given. It will work by keeping state with the previous study.
Also Read: How to setup a LAN connection in windows
Now any man is said, “Tahmid, see chair tables in front of you.” In this way no computer can be called, computer input has to be certainly in binary – because how much it tries to work like a man, it is not a man. We know that the switch is on or off for accepting each input of the computer. The neural network will give the answer only in the negative or positive way. Suppose you input by showing table chair’s picture,
- Is there a place to lean against the back of it?
- Has it’s surface?
- Is it fitted with soft mattress in it?
- Here can you sit comfortably for a long period of time?
- Can be kept a lot of things on it?
Now, the computer will answer “in the case of table” yes, no, yes, yes, no or in binary 01001. In another word at the time of learning, it will see the table as 01001 and chair as 10110, and it will also understand that is it a chair or table.
Last Word: If it is possible to make improve hopefully in a neural network then total computing technology will be changed. In spite of that, in present neural networks are being used in the work of aircraft autopilot, credit card transaction, radar scanning controlling, recognizing the handwriting, voice recognition, catching email spam. In the future, it will be possible to make an extremely intelligent robot, which will act exactly like humans. Anyway, really today’s matter was very complicated, sometimes some technical term came in it but after that also I have tried of my level best to reveal it in easy language.
I hope, you have liked it extremely- I like to learn new things so much. Comment us below if you have any question about today’s matter and share this article definitely.
One time this technology will improve so much that computer will work on controlling world and computer will get out of control of the man, the talent will be zero of all because everybody will keep information in computer and that’s why the average of education will decrease. At last higher civilization will be destroyed only for computer. We should aware of it now.