In supervised learning, ANNs are fed labeled datasets that predict the correct answer in advance. For example, training data for facial recognition might involve labeling each image with different terms related to emotions. After the network is trained, it will begin to “predict” the emotion of a new image of a human face that it has never processed before.
With unsupervised learning, the ANN is trained with unlabeled data that gives no clues about what it is looking at.
This type of ANN, in particular, is also called a black box because it is unclear how the ANN “learns”—how all the individual neurons in the ANN work together to arrive at the final output. While it is clear how an individual node applies a given weight during training and how it computes an output, it is less easy to understand how multiple nodes or “neurons” arrive at the final output.
In other words, studying and analyzing the structure of some types of ANN does not provide complete insight into its performance and how it arrives at its predictions. Hence the term “black box”.
Some models of artificial neural networks
Numerous types of artificial neural network models are used in supervised and unsupervised learning. These differ in terms of complexity, use cases and structure. Some of the main ones are Perceptron, Multilayer Perceptron Neural Networks, Feedforward Artificial Neural Networks and Radial Basis Function Artificial Neural Networks. An additional model that can be used in both supervised and unsupervised learning is a recurrent neural network.
A feedforward artificial neural network is one in which data flows unidirectionally between input and output nodes. Although there can be many different layers with many different nodes, the data moves in only one direction. Feedforward models are mainly used for simple classification problems.
Radial basis function (RBF) networks use activation functions for estimation. It is a mathematical function that calculates the absolute value between the centroid and a given point. They are similar to MLP, but use only one hidden layer instead of multiple layers – they usually have an input layer, a layer with radial basis function nodes with different parameters, and an output layer.
A common application of radial basis function neural networks is in system control, for example, systems that control power restoration after a power cut by prioritizing the repair of a large number of people.
A recurrent neural network processes data sequentially. This model of ANN will feed the data and loop back to previous steps in the artificial neural network.
Hidden layers are recursive, allowing data to be looped and retained. The output is then looped back into the input, improving development for the next input. This allows him to predict better and achieve the task efficiently.