Neural Nets

Hey guys. We are all definitely coming up with a word Neural Nets once in our daily life. So, What is Neural Nets? Why is it called so? What does it do? Why is it so popular among other ML algorithms? Come on, Let's know them in this post.

What is Neural Nets?

Neural Nets (Neural Networks) is a powerful algorithm for solving Real-Time problems easily. The Word Neural Networks is inspired by the Neurons in our brain. Neural Nets work the as same as the function of Neurons. Neurons contain axon (tail) and dendrite (head) which acts as the ending and starting point of a neuron. Every dendrite of a neuron connects to the axon of another neuron. So, by connecting multiple neurons together, a network will be formed and it is used to transfer the electric signals to a destination in our body.

Neural Nets works as the same as Neurons in our body. Neural Nets also have neurons that are formed to be a Layer of Neurons. The Intermediate layers are called Hidden Layers. So, One Input layer, One Intermediate layer, and One Output layer forms a Simple Neural Network. If there are two or more hidden layers, then it is called Deep Neural Network, from which the term Deep Learning comes from.

How does it work?

Neural Network works on the basis of simple mathematical formula which we had studied in our lower classes.

y = mx + c

Yes..!! You are right. It is the formula for a line having slope in it. But, How does it related to Neural Networks? Let me explain to you.

There are many types of Neural Networks. We'll discuss in general, not so specific. Neural Nets work by tuning the weights of the Nodes (Neurons) to get the accurate output (prediction). We'll be adding Bias at each node by the beginning of each propagation. So what Bias does is, it helps the model to be flexible and without bias, all the points will flow through the origin and it doesn't make a good sound in Neural Networks. Because, when the lines are flowing through the origin, we have more similar lines overlapping together, results in bad predictions.

So, the formula for the input of an intermediate node can be written as,

node = (weight)*node_value + bias

This is the simple working of Neural Nets. We'll see about their types in upcoming articles.

Why is it so popular?

Though there are other algorithms out there that can solve the problems which Neural Nets solves, most ML / DL engineer / Data Scientist prefers the Neural Nets. Because it is more accurate that the other algorithms and faster than others out there. It can handle large amounts of data easily. Development in processing capability gives Neural Network the special power to rule the AI Industry.

Most famous methods like Image Processing, Audio Processing using Neural Networks are very accurate compared to the native implementation of these methods using a library like OpenCV, etc. Introduction of Tensorflow in Deep Learning, Introduction of data handling and visualizing libraries like NumPy, Pandas, matplotlib, Seaborn provides a big strength to work together in a single platform and perform many tasks without any conflicts. Compared to any other algorithms, Neural Networks are easy to manage and we can tune the model using hyperparameters which we will see in upcoming articles.