Artificial Neural Network – 101
You must have heard people talking about Artificial Intelligence, Deep Learning revolutionizing the world. Applications getting better and better in Speech Recognition, Facebook getting better in recognizing the face in an image and accurately tagging them but, have you ever wondered how they are able to do these cool stuff? What is the technology behind them?
The answer to all the questions is Artificial Neural Network…….
Another question that follows the previous one is, what is Artificial Neural Network(ANN)?
Here is how Wikipedia and other articles defined it:
Artificial neural networks (ANNs) or connectionist expert systems are computing systems inspired by the biological neural networks that constitute animal brains. Such systems learn to do tasks by considering examples, generally without task-specific programming. ANNs are considered non-linear statistical data modeling tools where the complex relationships between inputs and outputs are modeled or patterns are found.
Didn’t get it??????
Don’t worry, I too scratched my head the first time I read it!!!
Let’s Start Fresh and Simple….
Before starting with Neural Networks let’s go to our school days, pull out your Biological Science book, open chapter Neuron
Image of a Neuron
You will see this image of a Neuron which is the major component of our brain responsible for transmitting and processing information through electrical and chemical signals. Millions of neurons connect to each other forming a network of neurons that help in coordinating human body day-to-day tasks.
Artificial Neural Network is inspired by the same idea of biological neural networks. Instead of neurons, we have processing units/computing systems also called Perceptron which, is responsible for processing mathematical algorithms on inputs and provide optimum output.
Simplest Artificial Neural Network looks like this:
ANN is based on the collection of many interconnected artificial neurons, which receive input, modify their internal state based on inputs and produce Output depending upon the input and activation algorithm.
Each connection between neurons is a process of transmission of information or we can also say each neuron feed information to another connected neuron, this transmission of information is based on Feedforward algorithm (we’ll talk about it later).
How this Process actually works?
The whole process of the neural network is a bit typical to implement, so first, let’s get the high-level understanding of the steps involved in creating a neural network:
- The First step is to provide inputs to perceptron with some weights for which there is known-answer. Weights of the input denote how much the input matters in the model.
- Calculate the sum of all the inputs multiplied by their weights i.e.
X1×W1 + X2×W2
Where X= Input; W=Weight of the Input
- Ask the perceptron to guess the answer.
- Computation of output(Y) using Activation function which allows you to confirm the output to some desired range.
- Adjust all the weights according to the error using weight adjustment method i.e.
W1=W1 + ΔW1 (Change in W1)
- Return to step 1. And repeat.
So, I hope that was simple enough to get the basic idea about Neural Networks. In the next blog, we’ll try to deep dive into neural networks and implement a simple example create our own neural network.
Till then stay updated and Happy Deep Learning!!!!!!