Cube Kohonen SelfOrganizing Map CKSOM Model With New
Cube Kohonen Self-Organizing Map (CKSOM) Model With New Equations in Organizing Unstructured Data Presenter : YU-TING LU Authors : Seng Poh Lim and Habibollah Haron 2013. TNNLS Intelligent Database Systems Lab
Outlines n Motivation n Objectives n Methodology n Experiments n Conclusions n Comments Intelligent Database Systems Lab
Motivation • For unstructured data, there is no connectivity information between data points. As a result, incorrect shapes will be obtained during the imaging process. • 2 -D Kohonen maps are limited because they are unable to cover the whole surface of closed 3 -D surface data. Intelligent Database Systems Lab
• closed surface • open surfaces Intelligent Database Systems Lab
Objectives • The aim of this paper is to use KSOM to organize unstructured data for closed surfaces. • Enhancements to the KSOM for organizing unstructured data for closed 3 -D surfaces and solving the problems of 2 -D and 3 -D KSOM. Intelligent Database Systems Lab
Methodology Intelligent Database Systems Lab
Methodology – Acquiring data • Talus bone data • 5, 235 points. Intelligent Database Systems Lab
Methodology – Acquiring data Intelligent Database Systems Lab
Methodology – Initializing parameters Intelligent Database Systems Lab
Methodology – Merging neurons Intelligent Database Systems Lab
Methodology – Merging neurons Intelligent Database Systems Lab
Methodology – Merging neurons Intelligent Database Systems Lab
Methodology – Merging neurons Intelligent Database Systems Lab
Methodology – Detecting neighbors Intelligent Database Systems Lab
Methodology – Generating weights, learning process and producing output Intelligent Database Systems Lab
Experiments - Analysis and validation of images Intelligent Database Systems Lab
Experiments - Analysis and validation of images Intelligent Database Systems Lab
Experiments - Analysis and validation of metric evaluation Intelligent Database Systems Lab
Experiments - Analysis and validation of equations Intelligent Database Systems Lab
Experiments - Analysis and validation of equations Intelligent Database Systems Lab
Quantization errors =0. 0001 Quantization errors =0. 00007 Intelligent Database Systems Lab
Conclusions • The model solved 2 -D KSOM problems by covering the whole surface of a closed surface and handled connectivity problems of 3 -D KSOM. • The model also contained fewer quantization errors compared to 2 -D and 3 -D KSOM. Intelligent Database Systems Lab
Comments • Advantages -Fewer quantization errors • Applications -Self-Organization Map -Organization medical image data Intelligent Database Systems Lab
- Slides: 23