【源码】POD - Proper Orthogonal Decomposition (Wrapper)

This function is just a nice wrapper for Matlab’s built-in SVD function, which only operates in 2D data sets. The wrapper just reshapes the matrices before and after applying SVD such that the input series X can be n-dimensional and ensures the output modes U are also n-dimensional.

=====

%POD has several names (POD, SVD, PCA, …) and is a mathematical tool that

%highlights the principal modes of a random series of variables (i.e.,

%financial measurements, fluid flow fields, structural vibrations, neuron

%firing patterns, video frames, and basically anything imaginable!)

%The POD tool is extremely useful in data analysis to identify the most energetic modes of a complex system by simply performing measurements when its operating. (But if you’re already here you already know all of this.)

%Inputs:=

%X - Time series or snapshots to be analyzed. Can be a n-dimensional matrix

%(n>=2). First dimension is time/snapshot, and all other dimensions are

%conserved for output.

%Outputs:=

%U - Mode matrix, has the same size of X. First dimension of X was

%“time/snapshot”, therefore first dimension of U will be “mode”. Modes are

%automatically organized from most energetic U(1,:,:…) to least

%energetic.

%S - Mode eigenvalues. S.^2 gives the mode energies. S is a diagonal

%matrix, so returns just the vector S to save one line of post-processing.

%V - Mode time-series or snapshot coefficients - V contains the

%contribution of each mode to a specific snapshot of X. USV gives back X

%(given U and X are reshaped into 2D matrices again).

更多精彩文章请关注公众号：