Artificial Intelligence and Deep Learning in Five Minutes

• Read original article here

I don’t want to explain it in the usual way of other experts, just come out of other definitions you read before, before reading this. Before that, why experts are calling it deep learning? Is it that much deep? you can presume whatever before knowing about this in detail. Presumptions may differ from person to person.

But what actually it is?

Here I shed light on this topic and this article is an attempt to explain this concept in a comprehensible way that even a non-technical person can understand.

Things you are going to learn through this article,

To understand what is deep learning, you should have known something about his grandfather ‘Artificial Intelligence’ and father ‘Machine Learning’ before. Because deep learning is something which is part of machine learning and machine learning is a part of artificial intelligence. Hope this below picture helps you to see the light.

If a human creates one which depicts a natural thing, we usually call it an ‘artificial’. Scientists were once fed up with this machine’s programmed activities and expected more from these to act like a human. Ordinarily, a machine cannot think by itself, but they wanted the machine to think by itself as a human to solve a lot of real-world problems. In this way, a machine can solve any problem even without human intervention or command. Another main advantage is machines are too faster and precise than human. So they decided to inject a human’s level of intelligence into machines and later succeeded in that with these techniques. This intelligence is exclusively made by humans for machines, so that’s how it became ‘Artificial Intelligence’, a man-made intelligence.

In this way, human’s intelligence can be called ‘natural intelligence’, because it has been given by nature.

The main aim of this artificial intelligence is to make a machine which can learn by itself like human beings. whenever humans learn something, initially they do some mistakes, learn from it and try to do that again without doing that same mistake again. In the same way, this machine learning and deep learning techniques work. These methods get trained with the help of data most of the times.

After 1980, machine learning started to gain traction, but it took 30 years for deep learning to that. we can say deep learning is an enhanced version of machine learning, but it has its own cons and pros. Deep Learning is a subset of machine learning. Machine learning is a good fit if the amount of training data is very small. Deep learning outperforms machine learning techniques always because of its architecture and supremacy. But both are the subsets of artificial learning.

As I already said, let’s know what this deep learning can actually do. Actually, it works more similar to human brains. Imagine a man practices juggling, whenever he misses the ball, every neuron from his brain count it as a mistake and transmit the signal within themselves and try to rectify that next time. This is how every human usually learns all the things and this is how our brain actually works when learning. In fact, our brain has around 100 billion neurons. They just exactly emulated the same human brain mechanism and made deep learning to work in that way. Neurons are connected within themselves and form a network called a neural network. Deep learning usually gets the input from the input layer, the input will flow through the many layers of the network to process and come up with an output. If the output is not as expected, it will count it as the mistake and again send the output to the input layer and make some changes inside the layer to rectify it. It will keep sending back till it gets the expected output. This is what we call it Back Propagation of Error.

Neural network is a hard worker who always works till he gets the perfection he wants.

So ‘learn from mistakes’ is not only for humans now.

This is not transmission line cables.

This is the architecture of an artificial neural network.

Why it is called Deep Learning?

There is no direct answer to this. We could see that, it learns the behaviour of the data and finds the patterns in it, so we can say that it keeps learning something from the data, so the word ‘learning’ is justified here. But why the word ‘deep’ is here? I was so confused and asked google about this. Many have said that it may be because of its dense hidden layers. We can specify the number of layers we want if we really want to increase the capacity of the model with fewer data. We can give it arbitrarily when we code.

Still, it is not deep? Just rotate the neural network 90 degrees rightwards to see its deepness, it increases with the number of layers. So the word ‘deep’ is also justified.

Why artificial intelligence is so popular today?

The reason for artificial intelligence boon is because of a lot of factors and I listed some below,

Deep learning is one of the members of the artificial intelligence family. There are some important architectures to do different tasks. Deep learning is efficient than machine learning in one aspect but it takes more time to get trained because of the involvement of lots of layers and data volume which makes it computationally cost.