Schedule for autumn 2000

Date Speakers Topics
Patrik Hoyer
Mika Inki
Session #1: Redundancy reduction and ICA
"Unsupervised Learning" & "Finding minimum entropy codes" (H Book Chapters 1 & 14)
ICA Tutorial (extra material)
Vesa Siivola
Jarkko Salojärvi
Session #2: Information maximization
"Local synaptic learning rules suffice to maximize mutual information in a linear network" (H Book Chapter 2)
"An information-maximization approach to blind separation and blind deconvolution" (H Book Chapter 8)
Jarkko Venna
Markus Koskela
Teemu Hirsimäki
Session #3: Image feature extraction by sparse coding and ICA
"What is the goal of sensory coding?" (H Book Chapter 7)
"The `independent components' of natural scenes are edge filters" (additional material)
"Sparse coding with an overcomplete basis set: A strategy employed by V1?" (additional material)
Jaakko Särelä
Jarmo Hiipakka
Session #4: Analyzing brain signals by ICA
"Searching for independence in electromagnetic brain waves" (G Book Chapter 10)
"Analysis of fMRI by blind separation into independent spatial components" (additional material)
Harri Kalaja
Yuan Zhijian
Jaakko Peltonen
Session #5: Text document analysis by ICA
"ICA on noisy data: A factor analysis approach" (G Book Chapter 11) [CANCELLED!]
"Independent components in text" (G Book Chapter 13)
ICA of text documents (TechRep by Mark Girolami)
Matti Aksela
Heikki Astola
Session #6: Learning invariant image features
"Learning invariance from transformation sequences" (H Book Chapter 5)
The adaptive-subspace self-organizing map (additional material)
Juha Karvanen
Janne Nikkilä
Session #7: Analyzing dependencies in ICA image features
Independent subspace analysis (additional material)
"The independence assumption: Analyzing the independence of the components by topography" (G Book Chapter 3)
Tapani Raiko
Antti Honkela
Session #8: Bayesian methods applied to ICA
Tutorial on Bayesian methods (additional material) & "Ensemble learning" (G Book Chapter 5)
"Bayesian non-linear independent component analysis by multi-layer perceptrons" (G Book Chapter 6)
Ville Viitaniemi
Paulo Esquef
Session #9: Coherence as a basis for learning
"Learning perceptually salient visual parameters using spatiotemporal smoothness constraints" (H Book chapter 6)
"Learning mixture models of spatial coherence" (H Book chapter 12)
Vuokko Vuori
Ramunas Girdziusas
Session #10: More on learning features
"Emergence of position-independent detectors of sense of rotation and dilation with hebbian learning: An analysis" (H Book Chapter 4)
"Independent component analysis of textures" (additional material)

NOTE: "H Book" refers to Hinton & Sejnowski (eds): "Unsupervised Learning",
while "G Book" refers to Girolami (ed): "Advances in Independent Component Analysis"