How artificial intelligence works and how to create it (part 1)

Artificial intelligence plays a very important role today.
We find it everywhere. Whether it's in the car on the cell phone or in our remote control.
But how it exactly works seems to be a mystery to many people.
But it is quite simple and highly exciting.
It's definitely worth taking a look at how it works.
In this series I would like to show you how an artificial neural network works and how you can create it.

The Neurons

The network consists of neurons. Neurons are cells that can perform certain calculations.
These neurons are interconnected and together they form a network.
We distinguish three types of neurons: Input neurons, hidden neurons and output neurons.

This is an input neuron. The only function it has is to take in and pass on values from outside.
It does not perform any kind of calculation.

The hidden neuron is a little more complicated. It adds up all the received values and a so-called activation function is applied to the sum. The result of that will be transfered to the following neurons.
The activation function could be, for example, the piecewise linear function, where above a certain threshold value the result is linear to the input.

This has the advantage that inaccurate data that is below the threshold is not allowed to pass through and so it does not disturb the network.
There are also other activation functions. Most commonly used is the sigmoid function

The last neuron, the output neuron, also adds up all received values and passes the sum to an activation function. But this time the result is output and not passed to another neuron.

These types of neurons form the input-layer, hidden-layer and output-layer.
The value of each input neuron is passed to each neuron in the hidden-layer. The results of the neurons in the hidden-layer are passed to the next neurons.
You can now imagine it like this: A network should identify animals in pictures. Now each input neuron could take the value of one pixel.The output neurons could output the response as binary code, where each output neuron represents a binary digit. So for example "cat" could come out.
How this is really connected and how the network knows the result you will learn tomorrow in the second part.
If you want to learn more about this topic or other IT-topics you can subscribe to me.
I hope you enjoyed it.