When compared to a more traditional classifier such as a logistic regression model, or even a tree-based model such as random forest or a gradient boosting machine, deep neural networks have a few nice advantages.

As with the regression we did in Chapter 2, Using Deep Learning to Solve Regression Problems, we don't need to select or screen features. In the problem that we have selected in this chapter, there are 178 input variables. Each input variable is a specific input from an Electroencephalogram (EEG) labelled x1..x178.  Even if you were a medical doctor, it would be difficult to understand the relationship between that many features and the target variable. There is a good chance that some of those features are irrelevant, and a better chance that some higher-level interactions might exist between those variables and the target. If using a traditional model, we'd get the best model performance if we went through a feature selection step. That's not needed when using deep neural networks.