VS298: Natural Scene Statistics

From RedwoodCenter
Revision as of 07:01, 3 February 2014 by Bruno (talk | contribs)
Jump to navigationJump to search

This seminar will examine what is known about the statistical structure of natural visual and auditory scenes, and theories of how sensory coding strategies have been adapted to this structure.

Instructor: Bruno Olshausen

Enrollment information:

VS 298 (section 4), 2 units
CCN: 66489

Meeting time and place:

Monday 6-8, Evans 560

General reading:

  • Natural Image Statistics by Hyvarinen, Hurri & Hoyer
  • Geisler WS (2009) Visual perception and the statistical properties of natural scenes, Annual Review of Psychology paper

Schedule:

Date Topic/Reading Presenter
Feb. 3 Redundancy reduction, whitening, and power spectrum of natural images
  • Barlow: Theory of redundancy reduction paper
  • Atick: Theory of whitening paper
  • Field: 1/f2 power spectrum and sparse coding paper

Anthony DiFranco
Dylan Paiton
Michael Levy

Feb. 10 Whitening in time and color; Robust coding
  • Dong & Atick: spatiotemporal power spectrum of natural movies
  • Ruderman: statistics of cone responses
  • Karklin & Simoncelli: noisy population coding of natural images
Feb. 17 ** Holiday **
Feb. 24 ICA and sparse coding
  • Barlow (1972)
  • Olshausen & Field (1997)
  • Bell & Sejnowski (1997)
March 3 Statistics of natural sound and auditory coding
  • Clark & Voss: '1/f noise and music'
  • Smith & Lewicki: sparse coding of natural sound
  • Klein/Deweese: ICA/sparse coding of spectrograms
March 10 Higher-order group structure
  • Geisler: contour statistics
  • Hyvarinen: subspace ICA/topgraphic ICA
  • Lyu & Simoncelli: radial Gaussianization
March 17 Energy-based models
  • Osindero & Hinton
  • Roth & Black
March 24 Learning invariances through 'slow feature analysis'
  • Foldiak/Wiskott: slow feature analysis
  • Hyvarinen: 'Bubbles'
  • Berkes et al.
March 31 ** Spring recess **
April 7 Manifold and Lie group models
April 14 Hierarchical models
  • Shan & Cottrell: stacked ICA
April 21 Deep network models
April 28
May 5
May 12