Going through the basic theory of DL
As briefly described in Chapter 1, Effective Planning of Deep-Learning-Driven Projects, DL is a machine learning (ML) technique based on artificial neural networks (ANNs). In this section, our goal is to explain how ANNs work without going too deep into the math.
How does DL work?
An ANN is basically a set of connected neurons. As shown in Figure 3.1, neurons from an ANN and neurons from our brain behave in a similar way. Each connection in an ANN consists of a tunable parameter called the weight. When there is a connection from neuron A to neuron B, the output of neuron A gets multiplied by the weight of the connection; the weighted value becomes the input of neuron B. Bias is another tunable parameter within a neuron; a neuron sums up all the inputs and adds the bias. The last operation is an activation function that maps the computed value into a different range. The value in the new range is the output of the neuron, which gets passed...
