Nonlinear Dimensionality Reduction Approach (ISOMAP) 2006. 2. 28 Young Ki Baik Computer Vision Lab. Seoul National University
Dec 13, 2015
Nonlinear Dimensionality Reduction Approach
(ISOMAP)
2006. 2. 28Young Ki Baik
Computer Vision Lab.
Seoul National University
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
References
A global geometric framework for nonlinear dimensionality reduction
J. B. Tenenbaum, V. De Silva, J. C. Langford (Science 2000)
LLE and Isomap Analysis of Spectra and Colour Images
Dejan Kulpinski (Thesis 1999)
Out-of-Sample Extensions for LLE, Isomap, MDS, Eigenmaps, and Spectral Clustering
Yoshua Bengio et.al. (TR 2003)
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Contents
Introduction
PCA and MDS
ISOMAP
Conclusion
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Dimensionality Reduction
The goalThe meaningful low-dimensional structures hidden in their high-dimensional observations.
Classical techniquesPCA (Principle Component Analysis)
– preserves the variance
MDS (MultiDimensional Scaling)
- preserves inter-point distance
ISOMAP
LLE (Locally Linear Embedding)
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Linear Dimensionality Reduction
PCA
Finds a low-dimensional embedding of the data points that best
preserves their variance as measured in the high-dimensional
input space.
MDS
Finds an embedding that preserves the inter-point distances,
equivalent to PCA when the distances are Euclidean.
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Linear Dimensionality Reduction
MDS
distances
Relation
ijd
)()( 2ji
Tjiij xxxxd
221 ijdA
matrix centering theis H , HAHB
)()( xxxxb jT
iij T
T XX(HX)(HX)Bthen
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Nonlinear Dimensionality Reduction
Many data sets contain essential nonlinear structures
that invisible to PCA and MDS
Resort to some nonlinear dimensionality reduction
approaches.
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
ISOMAP
Example of Non-linear structure (Swiss roll)
Only the geodesic distances reflect the true low-dimensional
geometry of the manifold.
ISOMAP (Isometric feature Mapping)
Preserves the intrinsic geometry of the data.
Uses the geodesic manifold distances between all pairs.
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
ISOMAP (Algorithm Description)
Step 1
Determining neighboring points within a fixed radius based on
the input space distance .
These neighborhood relations are represented as a weighted
graph G over the data points.
Step 2
Estimating the geodesic distances between all pairs of
points on the manifold by computing their shortest path
distances in the graph G.
Step 3
Constructing an embedding of the data in d-dimensional
Euclidean space Y that best preserves the manifold’s geometry.
jid ,X
jidG ,
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Step 1
Determining neighboring points within a fixed radius based on
the input space distance .
# ε-radius # K-nearest neighbors
These neighborhood relations are represented as a weighted
graph G over the data points.
ISOMAP (Algorithm Description)
jid ,X
ε
K=4
ij
k
jid ,X
kid ,X
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
ISOMAP (Algorithm Description)
Step 2
Estimating the geodesic distances between all pairs of points on the manifold by computing their shortest path distances in the graph G.
Can be done using Floyd’s algorithm or Dijkstra’s algorithm
jidG ,
)},(),( ),,(min{),(
N1,2,...,k
othewise ),(
ji, gneighborin ),(),(
jkdkidjidjid
for
jid
jidjid
GGGG
G
G
ij
k jkdG , kidG ,
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
ISOMAP (Algorithm Description)
Step 3
Constructing an embedding of the data in d-dimensional
Euclidean space Y that best preserves the manifold’s geometry.
Minimize the cost function:
)()()(
),(),(
),(
12.121
NN
GG
jiY
IDID
and
jidjiD
yyjiDwhere
2)()(LYG DDE
Solution: take top d eigenvectors of the
matrix )( GD
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Experimental results
# FACE # Hand writing
: face pose and illumination : bottom loop and top
arch
MDS : open triangles
Isomap : filled circles
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU
Discussion
Isomap handles non-linear manifold.
Isomap keeps the advantages of PCA and MDS.
Non-iterative procedure
Polynomial procedure
Guaranteed convergence
Isomap represents the global structure of a data set within a single coordinate system.
Nonlinear Dimensionality Reduction Approach (ISOMAP)
Computer Vision Lab. SNU