Hi everyone, I am back again with an interesting topic. Can machines be able to think like humans and how we can create an artificial neural network to make machines smarter so that they can make decisions by themselves. For that, we need to study about how humans brain works?
Humans brain seems to be the most powerful tool on this planet for learning things and applying them. But why humans brain is the most powerful tool for learning and adapting things? The answer is there are 100 billions of neurons inside the human brain that are interconnected with each other which transmit signals with each other within a seconds and we can make decisions. This is how Neuron looks like-:
Well, the key point here is to understand that neurons by themselves are pretty much useless. It's like an ant and ant on its own can't build the anthill. But at the same time when we have a large number of ant ants, let's say about millions of ants they can build an anthill. The same thing with neuron by itself it's not that strong but when you have lots of neurons they work together to do magic. And how do they work together?
Well, that's what dendrites and axons are for. The dendrites are kind of like the receiver of the signal for the neuron and axon is the transmitter of the signal for the neuron, so this is how the neurons receive the signal and transmit it. The question is how we recreate this on a machine. Since the whole purpose of Deep learning is to mimic how the human brain works in the hope that by doing so, we are going to create something amazing infrastructure for machines to able to learn.
How we can create an Artificial Neural Network and how do Neural Network works?
Let's take an example, Suppose we are looking for a property, so let's say we have some inputs parameter, we have an area in square feet, we have several bedrooms, the distance to the city in miles and the age of a property and all of four are comprise are input layers. There's probably way more parameters that define the price of a property but for simplicity sake, we're going to look at just these four. Now in its basic form, the neural network only has an input layer and an output layer, no hidden layer, and an output layer is a price that we are predicting. In this form, what these input variables would do is they just weighted up by synapses and then the output layer would be calculated or basically, the price would be calculated and again here you can use any function we could use any of the activation function, could use logistic regression, squared function, but the point is that you get a certain output. Moreover, most of the machine learning algorithms that exist can be represented in this format.
But in neural networks, what we do have is an extra advantage that gives us lots of flexibility and power which is where that an increase in accuracy comes from, and the power is hidden layers. Now we have to understand how hidden layers give us extra power. To understand it lets get back on an example of a property, we have four parameters and as we know all inputs are not important to the neuron some are zero and some are non zero.
Let us suppose area and distance to the city are important to the neuron whereas bedrooms and the age of the property are not that important and another neuron or for hidden layer area, age and bedrooms are important but the distance to the city is not that much important. We are supposing that this is trained up a data-set. Question is why it is important so suppose it can be because in that city where the neural network has trained perhaps from there a lot of families with kids, with two or more children who are looking for properties with lots of bedrooms which are new. Hence, from the training that this neural network has undergone, it knows that when there's a property with a large area and have lots of bedrooms that are valued so that neuron has picked that up it knows that okay, so this is for I am gonna looking for I don't care about the distance to the city in miles, as soon as that criteria are met that neuron fires up. Like we have to take ant example that a single ant cant built the anthill but more and more ant can. And that's the situation where each one of the neurons by itself can't predict the price but together they have a superpower and can predict the price this is how the neural network works. We have just learned how neural network works .
This is all about this article of learning Neural Network as soon I learn different concepts I'll keep posting. So, there are more to go as I am learning.
Bye-Bye, See You in my next article. Until then enjoy Machine Learning and PEACE OUT.