NOTES: MAA Talk, San Diego, 2002

• Set 1: Pages 1-4
• Original Clown image + singular values
• First four single rank approximations
• Rank k approximations
• Set 2: Pages 5-8 (Orig)
• Sample Eigenface Script
• Eigenfaces: Sample Images
• Eigenfaces: Mean and Eigenfaces
• Eigenfaces: Two-dimensional rep
• Set 3: Page 10: Space and Time
  NOTE: The movie data and script file will be available shortly...
• Space 1: Sample movie images
• Space 2: Spatial mean and eigenvectors
• Space 3: Temporal eigenvectors
• Space 4: 3-d reduction of data
• Set 4: The QSVD and MNF
• Matlab Script for this example
• Example: Max Noise Fraction A noisy circle, Original Data
• The Noisy Circle: PCA Coordinates
• The Noisy Circle, Denoised
• Set 5: Signal Separation
• Matlab script to do this experiment
• The Original Mixture of Sound
• Sound in PCA basis (Still mixed)
• Sound in Quotient SVD Basis (Unmixed)

Other Links and Notes:

• A good reference: Kirby, M. and Sirovich, L. "Application of the Karhunen-Loeve procedure for the characterization of human faces", IEEE transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol. 12, No. 1 (Jan. 1990).
• Kirby's book goes into some detail into the "eigenfaces": Michael Kirby, Geometric Data Analysis, John Wiley and Son, 2001.
• Face Recognition Homepage
• MIT's Media Lab VISMOD homepage