Perceptron in Machine Learning

Machine Learning is becoming a very important part of the technology industry day by day. Perceptron is one of the key things of Machine Learning. Mr. Frank Rosenblatt invented this thing in the mid of 19^th century. In this article, we are going to explore perceptron in detail.

What is the Perceptron in Machine Learning?

It is supervised learning technology. It is mainly considered a unit of artificial neural networks. Perceptron is considered the simplest neural network structure. It has four different parameters. They are input values, weights and Bias, net sum, and an activation function.

Some basic things in perceptron

1. Perceptron inputs: These inputs can also be called input nodes. This is the primary level of the perceptron. It takes inputs. All the inputs will be in numerical value.
2. Weight: It shows the strength of each node in the perceptron.
3. Values: The value which every node takes is called the value.
4. Activation function: This function is a major part of the perceptron. This part of the perceptron decides whether the neural network will proceed or not. Activation functions can be of different types sign function, sigmoid function, and step function.

A basic example of perceptron

Let's understand this concept by a simple example. As we know, that perceptron works as a neural network. So we can think about our brain. When our brain decides something, it judges some point and then decides what to do. The same case happens with perceptron. Just think we are making a perceptron so that we can decide whether to go to watch a cricket match or not. This thing will surely depend upon some parameters like where the ground is, how it will cost, which teams will play, whether any friend is going or not and etc.

Criteria	Input value	Weight
The ground is near	X1 = 1	W1 = 0.4
The team is India	X2 = 1	W2 = 0.6
The cost is low	X3 = 0	W3 = 0.5
Virat Kohli is playing	X4 = 0	W4 = 0.4
It is a T20 match	X5 = 1	W5 = 0.6
It is an IPL match	X6 = 0	W6 = 0.2
Friends are going	X7 = 1	W7 = 0.6

Algorithm to calculate the result

Frank Rosenblatt suggested the algorithm as follows:

1. Set a threshold value.
2. Multiply all inputs with their weights.
3. Sum all the results.
4. Activate the output.

Set a threshold value

So, for this example, we are going to set a threshold value. But you have to remember that this value may vary from person to person. What I am trying to say is that if you are taking the threshold as two, then another guy may be your friend can take 1 or 3. If someone is taking 3 means, he wants genuine reasons to go to watch the game. If someone is taking 1 means, he wants very few reasons to go to watch the game.

Multiply all inputs with their weights

In this step, we will multiply values with weight. We have to find  X*W for every node.

1. X1 * W1 = 1 * 0.4 = 0.4
2. X2 * W2 = 1 * 0.6 = 0.6
3. X3 * W3 = 0 * 0.5 = 0
4. X4 * W4 = 0 * 0.4 = 0
5. X5 * W5 = 1 * 0.6 = 0.6
6. X6 * W6 = 0 * 0.2 = 0
7. X7 * W7 = 1 * 0.6 = 0.6

Sum all the results

In this step, we will sum all the results we got from the previous step. So the sum will be as follows.

 Sum= 0.4 + 0.6 + 0 + 0 + 0.6 + 0 + 0.6 = 2.2                            

Activate the output

Now, we will go through activate function, which will check whether the result exceeds the threshold or not. In this example result sum = 2.2, and our threshold value is 2. So we can conclude with the result that we can go to watch cricket match.

Here is the basic code for this example:

const threshold = 1.5;
const inputs = [1, 1, 0, 0, 1, 0, 1];
const weights = [0.4, 0.6, 0.5, 0.4, 0.6, 0.2, 0.6];


let sum = 0;
for (let i = 0; i < inputs.length; i++) {
  sum += inputs[i] * weights[i];
}


const activate = (sum > 2);