Perceptron in Deep learning

An introduction to Perceptron

What is Perceptron?

Perceptron is considered the building block of Deep Learning. You can say Perceptron is the simplest Neuron available. It's an algorithm that is used for Supervised machine learning. It can also be compared to a mathematical function.

A simple Perceptron

The above figure represents a simple perceptron. A perceptron is a linear classifier that is used is supervised classification. Perceptron has 3 parts.

1. Weight and bias: In this above expample w1 and w2 are the weight while l is the bias value. Usually whenever we perform linear classsification we use the mathematical formula xw1+yw2+l. So we denote these value in the given equation to find if the result is a positive value or a negetive value.
2. Sum: This is the second parameter in a perceptron. You can find the summation in the middle of the diagram. It takes all the values and gives the net result which can be a positive or a negetive value.
3. Function: The last one in a perceptron is a activation function. It can be any activation function which provides the range of value.

Types of perceptron Models:

1. Single pereceptron model.
2. Multi layer perceptron model.

Even though perceptrons are not complex and can be used widely but they come with some major limitations.

1. The output of the peceptron is always 1 or 0 as it has activation function at the end.
2. It can only classify linear seperable set of inputs. If the inputs are non linear the classification wont be proper.