Deep Learning with PyTorch Quick Start Guide

Principle component analysis

Another unsupervised method, often used in conjunction with supervised learning, is principle component analysis (PCA). This is used when we have a large amount of features that may be correlated and we are unsure of the impact each feature has in determining a result. For example, in weather prediction, we could use each meteorological observation as a feature and feed them directly to a model. This means the model would have to analyze a large amount of data, much of it irrelevant. Further, the data may be correlated so that we need to consider not just individual features but how these features interact with each other. What we need is a tool that will reduce this large amount of possibly correlated and redundant ...