The possibilities of a neural network may seem boundless, but there are in fact limitations as to what neural networks and machine learning in general can achieve.
First of all, neural networks have poor interpretability. In other words, neural networks often function as black-box algorithms, and it is difficult to interpret the results produced by a neural network. Take for example our project in Chapter 2, Predicting Diabetes with Multilayer Perceptrons, where we used a neural network to predict patients at risk of developing diabetes. The neural network takes in input, such as blood glucose level, blood pressure, age, and so on, and outputs a prediction of whether the patient is at risk of developing diabetes. Even though the neural network is able to make such a prediction with high accuracy, we do not actually know what are the factors that...