cg
changeset 1:7487ad7f5d8f
.
author | bshanks@bshanks-salk.dyndns.org |
---|---|
date | Sat Apr 11 19:35:08 2009 -0700 (16 years ago) |
parents | 29eee29f9bc1 |
children | 83d03daa5d1c |
files | .hgignore grant.html grant.odt grant.pdf |
line diff
1.1 --- a/.hgignore Sat Apr 11 19:12:32 2009 -0700
1.2 +++ b/.hgignore Sat Apr 11 19:35:08 2009 -0700
1.3 @@ -14,6 +14,7 @@
1.4 *.aux
1.5 *.xref
1.6 *.lg
1.7 +*.4??
1.8
1.9 syntax: regexp
1.10 .*\#.*\#$
2.1 --- a/grant.html Sat Apr 11 19:12:32 2009 -0700
2.2 +++ b/grant.html Sat Apr 11 19:35:08 2009 -0700
2.3 @@ -108,12 +108,12 @@
2.4 to label more than a few genes. Therefore, we must select only a few genes as
2.5 features.
2.6 Principle 3: Use geometry in feature selection
2.7 - When doing feature selection with score-based methods, the simplest thing to
2.8 - do would be to score the performance of each voxel by itself and then combine
2.9 - these scores; this is pointwise scoring. A more powerful approach is to also use
2.10 - information about the geometric relations between each voxel and its neighbors;
2.11 - this requires non-pointwise, local scoring methods. See Preliminary Results for
2.12 - evidence of the complementary nature of pointwise and local scoring methods.
2.13 + When doing feature selection with score-based methods, the simplest thing to do
2.14 + would be to score the performance of each voxel by itself and then combine these
2.15 + scores (pointwise scoring). A more powerful approach is to also use information
2.16 + about the geometric relations between each voxel and its neighbors; this requires
2.17 + non-pointwise, local scoring methods. See Preliminary Results for evidence of
2.18 + the complementary nature of pointwise and local scoring methods.
2.19 Principle 4: Work in 2-D whenever possible
2.20 There are many anatomical structures which are commonly characterized in
2.21 terms of a two-dimensional manifold. When it is known that the structure that
2.22 @@ -121,6 +121,8 @@
2.23 the analysis algorithm to take advantage of this prior knowledge. In addition,
2.24 it is easier for humans to visualize and work with 2-D data.
2.25 Therefore, when possible, the instances should represent pixels, not voxels.
2.26 + Aim 2
2.27 + todo
2.28 3
2.29
2.30 Aim 3
2.31 @@ -173,6 +175,7 @@
2.32 conceivable that the methods we propose to develop could be used to suggest
2.33 modifications to the human cortical map as well.
2.34 Related work
2.35 + todo
2.36 Preliminary work
2.37 Justification of principles 1 thur 3
2.38 Principle 1: Combinatorial gene expression
2.39 @@ -198,8 +201,6 @@
2.40 useful.
2.41 The requirement to find combinations of only a small number of genes limits
2.42 us from straightforwardly applying many of the most simple techniques from
2.43 - the field of supervised machine learning. In the parlance of machine learning,
2.44 - our task combines feature selection with supervised learning.
2.45 __________________________
2.46 1“WW, C2 and coiled-coil domain containing 1”; EntrezGene ID 211652
2.47 2“mitochondrial translational initiation factor 2”; EntrezGene ID 76784
2.48 @@ -228,6 +229,8 @@
2.49 genes which (individually) best match area AUD, according to gradient similar-
2.50 ity. From left to right and top to bottom, the genes are Ssr1, Efcbp1, Aph1a,
2.51 Ptk7, Aph1a again, and Lepr
2.52 + the field of supervised machine learning. In the parlance of machine learning,
2.53 + our task combines feature selection with supervised learning.
2.54 Principle 3: Use geometry
2.55 To show that local geometry can provide useful information that cannot be
2.56 detected via pointwise analyses, consider Fig. . The top row of Fig. displays
3.1 Binary file grant.odt has changed
4.1 Binary file grant.pdf has changed