What is Deep Learning?

Deep learning works by using artificial neural networks, which are multi-layered structures that analyze data through layers of computations, allowing for complex patterns to be identified and accurate predictions to be made.

Let’s take the example of a child learning to recognize birds in images. The parent first points to the bird in the picture and makes the correct sound. The child then tries to mimic the sound, which is corrected by the parent until the child understands the word. This learning process is similar to how artificial neural networks operate.

An artificial neural network consists of several layers, including the input layer, the hidden layer, and the output layer, each with its own weights and variables. The nodes in each layer analyze different features of the image. For example, the lowest layer analyzes the contrast of black and white pixels, while the second layer distinguishes lines based on the data from the first layer. By computing non-linear functions across multiple layers, the output layer produces the final value, which is the image’s classification and recognition.

Since these types of neural networks require multiple layers of neurons for data processing, they need to be compared layer by layer to operate, which makes the computational complexity relatively high. This is why deep learning is associated with the concept of ‘depth’.