Connectionists: New paper: Separation Theorem for Independent Subspace Analysis and its Consequences

[Zoltan Szabo]

Dear Connectionist-ers,

We are pleased to announce our recently accepted paper in the journal
of Pattern Recognition entitled 'Separation Theorem for Independent
Subspace Analysis and its Consequences'.

The paper could be of interest to many of you because of the following
reasons:

1)As it has been demonstrated several times ICA (independent component
analysis) on natural images leads to image filters that resemble to the 
simple cells in the V1 visual cortical area of the brain. If one uses
ISA (independent subspace analysis) instead of ICA on natural images,
i.e, one allows dependencies between some of the components, then ISA
will provide independent subspaces that show phase- and shift-invariant
properties similar to complex cells. 

2)The relevance of ISA is also underpinned by several other brain
related/biomedical applications including fMRI and MEG data
processing, ECG and gene expression analysis, action recognition. 

3)The paper concerns one of the most fundamental, long lasting open
hypothesis of the ICA/ISA research.

Abstract: Independent component analysis (ICA) -- the theory of mixed,
independent, non-Gaussian sources -- has a central role in signal
processing, computer vision and pattern recognition. One of the most
fundamental conjectures of this research field is that independent
subspace analysis (ISA) -- the extension of the ICA problem, where
groups of sources are independent -- can be solved by traditional ICA
followed by grouping the ICA components. The conjecture, called ISA
separation principle, (i) has been rigorously proven for some
distribution types recently, (ii) forms the basis of the
state-of-the-art ISA solvers, (iii) enables one to estimate the unknown
number and the dimensions of the sources efficiently, and (iv) can be
extended to generalizations of the ISA task, such as different linear-,
controlled-, post nonlinear-, complex valued-, partially observed
problems, as well as to problems dealing with nonparametric source
dynamics. Here, we shall review the advances on this field.

The in press version is now available at:
"http://dx.doi.org/10.1016/j.patcog.2011.09.007".

Comments/feedbacks are also very welcome.

Best,

Zoltan, Barnabas ("http://nipg.inf.elte.hu/szzoli",
"http://www.cs.cmu.edu/~bapoczos/")