Order: The key role in machine learning

Abstract

Emphasizing on feature extraction and classification, but ignoring order importance, is a general phenomenon in machine learning. Here, we first review the views on order in cognitive science briefly. Furthermore, a simple experiment on Yale database in face recognition is done to show the importance of order in machine learning. In the experiment, one kind of training sample set is randomly selected in one person's samples, but the ordinal of samples for different person is the same, while another kind of training sample set is completely randomly selected. The recognition performance using principal component analysis (PCA) on the first kind of set is significantly better than that on the second kind of set. This indicates the order's key role in machine learning, and it gives a new look on performance evaluation of previously published machine learning algorithms as well.