structure preserving embedding
DESCRIPTION
Structure Preserving Embedding. Blake Shaw, Tony Jebara ICML 2009 (Best Student Paper nominee) Presented by Feng Chen. Outline. Motivation Solution Experiments Conclusion. Motivation. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/1.jpg)
Structure Preserving Embedding
Blake Shaw, Tony JebaraICML 2009 (Best Student Paper nominee)
Presented by Feng Chen
![Page 2: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/2.jpg)
Outline
• Motivation• Solution• Experiments• Conclusion
![Page 3: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/3.jpg)
Motivation
• Graphs exist everywhere: web link networks, social networks, molecules networks, k-nearest neighbor graph, and more.
• Graph Embedding– Embed a graph in a low dimensional space.– What used for? Visualization, Compression, and More.– Existing methods: Planar Embedding (no crossing
arcs); Spectral Embedding (preserve local distances); Laplacian Eigenmaps (preserve local distances).
![Page 4: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/4.jpg)
Problem Statement
• Given only the connectivity of a graph, can we efficiently recover low-dimensional point coordinates for each node such that these points can be easily used to reconstruct the original structure of the network?
![Page 5: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/5.jpg)
Optimization Formulation
![Page 6: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/6.jpg)
Justifications
• Why can preserve the local distances?
)(trmax 1)(,0 KAKtrK
![Page 7: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/7.jpg)
Justifications
• Interpret the slack variable
![Page 8: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/8.jpg)
Implementation
• It is about finding a positive semi-definite matrix that can maximize an objective function, subject to some constraints about the matrix.
• This category of optimization problem is called “Semi-definite Programming”
• Available MATLAB tools include CSDP and SDP-LR.
![Page 9: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/9.jpg)
Related Work
• Spectral Embedding– Find the d leading eigen-vectors of the adjacency matrix
A as the coordinates. • Laplacian Eigenmaps– Employ a spectral decomposition of the graph Laplacian
L=D-A, where D=diag(A1), and use the d eigenvectors of L with the smallest nonzero eigenvalues.
– Look weird? But this approach has the fundamental from manifold theories in the field of topology!!
![Page 10: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/10.jpg)
Experiments
![Page 11: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/11.jpg)
Experiment
![Page 12: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/12.jpg)
Conclusion
• Innovative idea!• Technically simple!
![Page 13: Structure Preserving Embedding](https://reader036.vdocuments.net/reader036/viewer/2022062410/568161b0550346895dd173aa/html5/thumbnails/13.jpg)
Thank You!
Any Question?