Graph Based Classification Using kNN Averaging and Graph Embedding Criterion

In the recent years, because technology has improved rapidly, the size of data such as digital photographs becomes very high. For time and computation efficiency, we need to extract some features from this high dimensional data. Hence, in this paper, a new dimensionality reduction algorithm is proposed to extract features for the classification purpose. The proposed method is based on graph embedding which is a general framework for describing many dimensional reduction methods. In this framework, similarity and penalty graphs are constructed based on data relations. These graphs characterize the statistical or geometric property of the data that should be kept or avoided during dimensionality reduction. Our proposed method constructs these two graphs on data and uses the averaging idea among neighbor vertices of the graphs to imply the compactness in each class of data while separating different classes. Obtained results show that the proposed method improves the accuracy of classification task on data such as face and digit images

کلیدواژه ها:

Dimensionality reduction ، graph embedding ، similarity graph ، penalty graph ، classification