Any neural network is an artificial neural network that is used to build Deep Learning models.
Artificial Neural Networks work on the basis of the structure and functions of a human brain. Like the human brain has neurons interconnected to each other, neural network systems additionally have neurons that are interconnected to each other in various layers of the system. These neurons are called nodes.
ANN consists of 3 main layers:
a) Input layer - Accepts inputs in several different formats
b) Hidden Layers - Performs all the calculations and manipulations to extract hidden features and patterns
c) Output Layer - Produces the desired output.
ANN takes inputs and calculates the weighted sum of the inputs and adds a bias. This calculation is represented in the form of a Transfer function.
This calculated weighted sum is passed an input to an activation function to generate the output. Activation functions decide whether a node should be fired or not. Only those which are fired make it to the output layer. There are different activation functions available that can be applied depending on the kind of the task you are performing.
Some advantages and disadvantages of the artificial neural network are:
Without it, we wouldn‚??t have scratched the surface of deep learning. Deep learning is nothing but an ANN with multiple hidden layers, and it is responsible for the rapid development that‚??s going on in the Machine Learning industry right now. Before Deep learning, we were not nearly as good at stuff like image classification and speech recognition as we are today. Just look at all the things around you that are powered by deep learning - your Fitbit, Siri, Google Home, Amazon Alexa, and so many more.
Secondly, ANNs provided us the first step towards AI by generating a model based on how our own human body learns. Through mimicking neuron interaction within the body, researchers about 20 years ago were actually able to conquer something that had never been done before. Before neural nets, there were very few, if at all, models that were actually trained on how our body learned.
Disadvantages - pretty much just overuse of it. Everyone‚??s trying to apply deep learning to everything now, even things that don‚??t require it. It‚??s leading to a huge misunderstanding of the whole field, where people just spout buzzwords like Tensorflow, Neural Networks, and Machine Learning but have no clue what they actually mean.
This hype is shunning some people out of the field because they think they can‚??t enter it without every computer science/math skill in the book.