Hands-On Neuroevolution with Python

Preprocessing data

Always examine the range of possible data inputs and check for outliers. If you find that the scale of one input parameter differs from another by an order of magnitude, you need to preprocess the input data samples. Otherwise, the input data features with higher magnitude will have such a significant impact on the training process that they will ultimately outweigh the contribution of the other input data features. However, the small signals produced by the data inputs with small magnitude are often crucial to finding a successful solution. Delicate input signals can characterize subtle but valuable traits in the underlying process.