intelligent database systems lab presenter: jheng, jian-jhong authors: hansenclever f. bassani,...

Download Intelligent Database Systems Lab Presenter: Jheng, Jian-Jhong Authors: Hansenclever F. Bassani, Aluizio F. R. Araujo 2015 TNNLS Dimension Selective Self-Organizing

If you can't read please download the document

Upload: timothy-caldwell

Post on 21-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

  • Slide 1
  • Intelligent Database Systems Lab Presenter: Jheng, Jian-Jhong Authors: Hansenclever F. Bassani, Aluizio F. R. Araujo 2015 TNNLS Dimension Selective Self-Organizing Maps With Time-Varying Structure for Subspace and Projected Clustering
  • Slide 2
  • Intelligent Database Systems Lab Outlines Motivation Objectives Related Work Methodology Experiments Conclusions Comments 2
  • Slide 3
  • Intelligent Database Systems Lab Motivation
  • Slide 4
  • Intelligent Database Systems Lab 4 Original SOM subspace clustering It was not designed to deal with subspace clustering.
  • Slide 5
  • Intelligent Database Systems Lab 5 DSSOM The xed topology of SOM and DSSOM requires deep knowledge of the date to be dened.
  • Slide 6
  • Intelligent Database Systems Lab 6 SOM-TVS However, certain adaptations to this approach are required for dealing with subspace clustering.
  • Slide 7
  • Intelligent Database Systems Lab 7 LARFDS SOM Important modications with respect to DSSOM.
  • Slide 8
  • Intelligent Database Systems Lab Objectives Important modications with respect to DSSOM. 8
  • Slide 9
  • Intelligent Database Systems Lab S UBSPACE AND P ROJECTED C LUSTERING
  • Slide 10
  • Intelligent Database Systems Lab S UBSPACE C LUSTERING
  • Slide 11
  • Intelligent Database Systems Lab P ROJECTED C LUSTERING Projected clustering seeks to assign each point to a unique cluster, but clusters may exist in different subspaces. The general approach is to use a special distance function together with a regular clustering algorithm.
  • Slide 12
  • Intelligent Database Systems Lab Generalized Principal Components Analysis (GPCA) Sparse Subspace Clustering (SSC) Adaptive Resonance Theory (ART) Local adaptive receptive eld self-organizing map for image color segmentation (LARFSOM) Related Work 12
  • Slide 13
  • Intelligent Database Systems Lab GPCA PCA GPCA GPCA is an algebraic geometric method for clustering data not necessarily in independent linear subspaces.
  • Slide 14
  • Intelligent Database Systems Lab SSC SSC is based on the idea of writing a point (x j ) as a linear or afne combination of neighbor data points. It uses the principle of sparsity to choose any of the remaining data points as a possible neighbor.
  • Slide 15
  • Intelligent Database Systems Lab Methodology
  • Slide 16
  • Intelligent Database Systems Lab
  • Slide 17
  • Slide 18
  • Slide 19
  • Slide 20
  • Slide 21
  • Slide 22
  • EXPERIMENTS
  • Slide 23
  • Intelligent Database Systems Lab
  • Slide 24
  • Slide 25
  • Slide 26
  • Slide 27
  • Slide 28
  • Conclusions The behavior of LARFDSSOM was shown to have led to a signicant improvement over DSSOM in a number of points. It does not need to know the exact number of clusters. Improvement entails the computational cost. Improvement is the clustering quality. 28
  • Slide 29
  • Intelligent Database Systems Lab Comments 29