Voxel selection using submodular function optimization

Post date: Feb 4, 2013 1:24:00 AM

The goal here is to find a “good” subset of voxels B in the brain that collec- tively/jointly forms a representational (dis)similarity matrix (RDM) R that is closest to the ideal sematic (dis)similarity matrix SS obtained from some crite- ria. For instance, SS can be obtained from WordNet, survey, Wikipedia corpus, etc. There are many possible criteria for selecting a good subset, for instance, minimum/maximum number of voxels to be included in the selected voxel subset and/or voxel contiguity in the brain (3D or cortical) space. More detail here.

There are lots of good resources here:

Carlos Guestrin's CMU page.

OPTIMIZING SENSING: FROM WATER TO THE WEB [pdf]

------ Feature selection using MI + submodular function optimization -----

Slides from Bilmes:

http://melodi.ee.washington.edu/~bilmes/ee595a_spring_2011/lecture1.pdf

Another tutorial on SFO:

http://www.di.ens.fr/~fbach/submodular_fbach_mlss2012.pdf

Submodular function page:

http://submodularity.org/icml08/

Tutorial slides on Submodular function optimization (ICML2008) by Carlos:

http://submodularity.org/submodularity-slides.pdf

SFO Toolbox:

http://jmlr.csail.mit.edu/papers/volume11/krause10a/krause10a.pdf

Recommended readings:

1. 10] A. Krause and C. Guestrin. Near-optimal nonmyopic value of information in graphical models. In UAI, 2005
2. Information gain in Naive Bayes models is submodular. Also gives (1-1/e) hardness of approximation result for information gain in such models.
3. [12] A. Krause, C. Guestrin, A. Gupta, and J. Kleinberg. Near-optimal sensor placements: Maximizing information while minimizing communication cost. In Proceedings of the Fifth International Symposium on Information Processing in Sensor Networks (IPSN), 2006
4. Optimizing monotonic submodular functions subject to communication constraints.
5. [13] A. Krause, B. McMahan, C. Guestrin, and A. Gupta. Selecting observations against adversarial objectives. In NIPS, 2007
6. Maximizing the minimum over a set of monotonic submodular functions, with applications to robust experimental design.
7. Submodular meets Spectral: Greedy Algorithms for Subset Selection, Sparse Approximation and Dictionary Selection