【字典学习 autoencoder】Dictionary Lea

Dictionary Learning Tools for Matlab

Contents on this page:

Relevant papers and links to other pages:

A brief introduction,

Sparse Approximation,

Dictionary learning,

MOD or ILS-DLA

K-SVD

ODL

RLS-DLA

Experiments from the RLS-DLA paper

Sparse representation of an AR(1) signal, dictionary size 16x32

Recovery of a known dictionary, dictionary size 20x50

More examples.

Image compression, ICASSP 2011 paper, dictionary size 64x440

Dictionary properties, SPIE 2011 paper, dictionary size 64x256

Files and details.

How to install and test the files.

Attached files.TheImage Compressing Tools for Matlabweb page.

ILS-DLA, the Iterative Least Squares Dictionary Learning Algorithm by Engan et al. ILS-DLA includes Method of Optimized Directions (MOD).

K-SVD, the K-SVD method for dictionary learning by Aharon et al.

RLS-DLA, the Recursive Least Squares Dictionary Learning Algorithm paper by Skretting and Engan.

ODL, the Online Dictionary Learning for Sparse Coding paper by Mairal et al.

SPAMS, the page for the SPArse Modeling Software by Mairal.

ThePartial Search, paper presented at NORSIG 2003, by Skretting and Husøy.

TheICASSP 2011 paper, "Image compression using learned dictionaries by RLS-DLA and compared with K-SVD" by Skretting and Engan.

TheSPIE 2011 paper, "Learned dictionaries for sparse image representation: Properties and results" by Skretting and Engan.

mpv2, The documentation for the Java package with files for Matching Pursuit and Dictionary Learning by Skretting.

You may also see Skretting'sPhD thesisfor more on Dictionary (called Frame in the thesis) Learning.

Michael Elad has done much research on Sparse Representations and Dictionary Learning, most of hispublicationsare availabel online.

I highly recommend Elad's (2010) book:"Sparse and Redundant Representations: From Theory to Applications in Signal and Image Processing"