A deep learning model is a highly non-linear model that has got multiple layers with multiple nodes acting in sequence to solve a business problem. Every layer has been assigned a different task.
For example, if we have got a face detection problem, hidden layer 1 finds out which edges are present in the image. Layer 2 finds out the combination of edges, which start taking the shape of eyes, a nose, and other parts. Layer 3 enables the object models, which creates the shape of the face. The following diagram shows the different hidden layers:
Here, we have got a logistic regression model, also known as a single layer neural network. Sometimes, it is also called the most shallow network. The second network that can be seen here has got a two-layer network. Again, it's a shallow network, but it's not as shallow as the previous one. The next architecture has got three layers, which is making things more interesting now. The network is getting deep now. The last architecture has got a six layer architecture, which is comprised of five hidden layers. The number of layers is getting even deeper.