1. Neural Network Homework Report: Clustering of the Self-Organizing Map Professor ： Hahn-Ming Lee Student : Hsin-Chung Chen M9315928 IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL.11, NO.3, MAY 2000

2. OUTLINE  INTRODUCTION  CLUSTERING  SOM CLUSTERING  EXPERIMENTS  CONCLUSION

3. INTRODUCTION  DATA mining processes problem definition data acquisition. data preprocessing and survey data modeling evaluation. knowledge deployment.  Self-organization map feature: Dimensionality reduction of unsupervised learning Can applied in deal huge amounts of sample The original data set is represented using a smaller set of prototype vectors not to find an optimal clustering but to get good

4. CLUSTERING  two main ways approaches hierarchical approaches  agglomerative algorithm ： bottom-up strategies to build a hierarchical clustering tree  divisive algorithm ： top-down strategies to build a hierarchical clustering tree partitive approaches  k-means  optimal clustering is a partitioning minimizes distances within maximizes distances between clusters

5. CLUSTERING(cont.)

6. SOM CLUSTERING  SOM training first to find the best matching unit (BMU) the prototype vectors are updated.

7.  The SOM algorithm characteristic ： applicable to large data sets. The computational complexity scales linearly with the number of data samples it does not require huge amounts of memory that basically just the prototype vectors and the current training vector.

8. EXPERIMENTS  Tools: SOM_ToolBox 2.0 :  Data set: clown.datclown.dat Data set ( “ clown.data ” ) consisted of 2220 2-D samples.  cluster with three subclusters (right eye)  spherical cluster (left eye)  elliptical cluster (nose)  nonspherical cluster (U-shaped: mouth)  large and sparse cluster (body)  noise.(such as black x)

9.  Methods and Parameters: Cluster step 1:  Training Parameters of the SOM's  Map size: 19x17 Initial Neighborhood Widths: Rough Phases σ1(0): 10 Fine-Tuning Phases σ2(0): 2 learning rates:(The learning rate decreased linearly to zero during the training) Rough Phases : 0.5 Fine-Tuning Phases 0.05 Cluster step 2:  Method: K-Means Using 100 Runs

11. Experimental Results

14. single linkage dendrogram of 323 SOM Map unit

15. SOM Map average linkage dendrogram of 323 SOM Map unit

16. complete linkage dendrogram of 323 SOM Map unit

20. Conclusion