Dimensionality Reduction(维度降低)
Data Compression(数据压缩):e.g.Reduce data from 3D to 2D;Reduce data from 2D to 1D
Data Visualization(数据可视化):一般将数据维度降低到2D或3D
Principal Component Analysis(主成分分析)
Principal Component Analysis problem formulation(主成分分析问题方程)
PCA algorithm
然后求解协方差矩阵,进一步获得Ureduce矩阵
Reconstruction from compressed representation(压缩数据还原重建)
Choosing the number of principal components(选择主成分数量K)
选择方法:
1.K值从1开始增加,循环计算PCA算法,直至满足要求(太过麻烦,可行性低)
2.使用协方差矩阵计算结果中的S矩阵
Advice for applying PCA(应用PCA的建议)
Bad use of PCA: To prevent overfitting(PCA不适用于减少过度拟合)