Implementation In FPGAs Of Principal Component Analysis

Along with constantly progresses of information technologies, people have to confront high dimensional data in multitude, which include much redundant information so that researchers are difficult to obtain relationships between the data. Meanwhile, it's a big waste of resources when the high dimensional data are stored, transmitted and processed.Data dimensionality reduction technology researches how to reduce the data dimension and extract the implicit and helpful information with losing information as little as possible. Principal Component Analysis (PCA) is a typical algorithm of data dimensionality reduction, which realizes the aim by eigenanalysis of matrix and mapping the original data to a linear subspace that contains most information. Therefore, the strongpoint of PCA is simple computation, and there's little of information losing.Chapter 1 presented research background and research significance first, then determined the content and aim of the research, illuminated the paper's configuration at last.Chapter 2 started with analysis to the birth and development status of data dimensionality reduction technology, then presented and reviewed several methods of data dimensionality reduction in common use.Chapter 3 clearly put forward the geometric significance of PCA in 2D space, then deducted PCA in theory, concluded the calculated steps and properties of PCA. Finally, three applications of PCA were introduced in different fields.Chapter 4 focused on the architecture of implementation in FPGAs of PCA. The experiment results indicate that the system could implement PCA algorithm to different number or dimension of data with small error rate, higher computing speed, stable clock frequency and small amount of resources.In the end, Chapter 5 illuminated the significance of the paper, and made a prospect of the future research.