Download presentation
Presentation is loading. Please wait.
Published byGriffin Floyd Modified over 8 years ago
1
Intelligent Database Systems Lab Presenter : Chuang, Kai-Ting Authors : Rodrigo T. Peres, Claus Aranha, Carlos E. Pedreira 2013, InfSci Optimized bi-dimensional data projection for clustering visualization
2
Intelligent Database Systems Lab Outlines Motivation Objectives Methodology Experiments Conclusions Comments
3
Intelligent Database Systems Lab Motivation The problem of data visualization consists of generating a bi-dimensional projection of a high- dimensional data set.
4
Intelligent Database Systems Lab Objectives We propose a new method to project n-dimensional data onto two dimensions, for visualization purposes. We apply Differential Evolution as a meta-heuristic to optimize a divergence measure of the projected data. This divergence measure is based on the Cauchy– Schwartz divergence, extended for multiple classes.
5
Intelligent Database Systems Lab Methodology-Framework Cauchy-Schwartz divergence measure Differential Evolution Data transformation
6
Intelligent Database Systems Lab Methodology
7
Intelligent Database Systems Lab Methodology-Cauchy-Schwartz divergence measure
8
Intelligent Database Systems Lab Methodology-Information Theoretic Learning (ITL)
9
Intelligent Database Systems Lab Methodology-Information Theoretic Learning (ITL)
10
Intelligent Database Systems Lab Methodology-Information Theoretic Learning (ITL)
11
Intelligent Database Systems Lab Methodology-Computational complexity of the Dcs
12
Intelligent Database Systems Lab Methodology-Differential Evolution
13
Intelligent Database Systems Lab Methodology-Data transformation
14
Intelligent Database Systems Lab Experiment setup Synthetic data sets – Initial conditions. – Robustness of the method to very noisy dimesions. Real world data sets – Pen Digits – Lung Cancer – Compares monocytes-related dendritic cells, plasmocytoid dendritic cells and B-lymphocytes. – Compares monocytes and neutrophils. – Compares plasmocytoid dendritic cells and neutrophils.
15
Intelligent Database Systems Lab Experiment-Kernel width
16
Intelligent Database Systems Lab Experiment-Synthetic data sets1
17
Intelligent Database Systems Lab Experiment-Synthetic data sets2
18
Intelligent Database Systems Lab Experiment-Real world data sets
19
Intelligent Database Systems Lab Conclusions Using this method, we promote the bi-dimensional visualization of high-dimensional data sets with optimized cluster separation.
20
Intelligent Database Systems Lab Comments Advantages – The method performed well. Disadvantages – It may be slower to train on data sets with a larger number of cases. Applications – Visualization.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.