0% found this document useful (0 votes)
96 views17 pages

Self Organized Maps: Inspire Educate Transform

The document discusses self-organizing maps (SOMs), an unsupervised machine learning technique invented by Teuvo Kohonen. SOMs arrange nodes in a grid and train them to cluster and visualize high-dimensional input data by adjusting the nodes' weight vectors. The document provides examples of using SOMs to analyze Irish census data and cluster localities based on attributes like education, household size, and car ownership.

Uploaded by

Sahil Goutham
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
96 views17 pages

Self Organized Maps: Inspire Educate Transform

The document discusses self-organizing maps (SOMs), an unsupervised machine learning technique invented by Teuvo Kohonen. SOMs arrange nodes in a grid and train them to cluster and visualize high-dimensional input data by adjusting the nodes' weight vectors. The document provides examples of using SOMs to analyze Irish census data and cluster localities based on attributes like education, household size, and car ownership.

Uploaded by

Sahil Goutham
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

Inspire…Educate…Transform.

Self Organized Maps

Lt. Suryaprakash Kompalli


Senior Mentor, International School of
Engineering

The best place for students to learn Applied Engineering http://www.insofe.edu.in


Self organizing maps

• Invented by Teuvo Kohonen, a professor of


the Academy of Finland.

• Kohonen described SOM as a visualization


and analysis tool kit for high dimensional
data.

CSE 7 30 5 c
• SOMS are however used for clustering,
dimensionality reduction and
classification.
The best place for students to learn Applied Engineering 2 http://www.insofe.edu.in
The Basics
http://ssdi.di.fct.unl.pt/aadm/aadm1011/slides/LectureSOM.pdf

Sample
Best matching unit
Original nodes
Nodes in weight spac
x1 x2 x3 ……. xd

• Several nodes arranged in a grid at uniform distance.


• Weight from each input (1…d) to node (1…n)

CSE 7 30 5 c
• Training modifies weights:
– Similar inputs activate neighboring nodes
– After training, distance between weights indicates how close the
nodes are

The best place for students to learn Applied Engineering 3 http://www.insofe.edu.in


BMU

• Which of the two nodes 1:(0.1, 0.2, 0.3) and 2:


(0.9, 0.1, 0.1) are closer to the input vector I=(1,
0, 0).
 
• Distance1 = sqrt( (1 - 0.1)2 + (0 - 0.2)2+ (0 -
0.3)2 ) = 0.96

CSE 7 30 5 c
Distance2= 0.17
  Hence, BMU is 2

The best place for students to learn Applied Engineering 4 http://www.insofe.edu.in


Training?

•  

CSE 7 30 5 c
The best place for students to learn Applied Engineering 5 http://www.insofe.edu.in
Training?
 

•  

CSE 7 30 5 c
The best place for students to learn Applied Engineering 6 http://www.insofe.edu.in
SOM Parameters
http://ssdi.di.fct.unl.pt/aadm/aadm1011/slides/LectureSOM.pdf

Classical vs Batch

Weights updated at each sample

Weights updated at each epoch

CSE 7 30 5 c
   

The best place for students to learn Applied Engineering 7 http://www.insofe.edu.in


How to use SOM?
Pre-process data: Generate
Standardization Build Self heatmaps and
Remove invalids Organizing Maps clusters to
Remove outliers interpret data

• In a random class, if students with similar


features sit as close as possible to each
other
– Age, grades, hobbies etc

CSE 7 30 5 c
• If you examine their grades with respect
to new positions, this will be a SOM

The best place for students to learn Applied Engineering 8 http://www.insofe.edu.in


SOMs in Practice

• Example from a tutorial by Shane Lynn


– Other examples:
• http
://www.r-bloggers.com/self-organising-maps-for-customer-segmentation-using
-r
/
• http://
manuals.bioinformatics.ucr.edu/home/R_BioCondManual#clustering_primer

• Census data from Dublin:

CSE 7 30 5 c
– ~4000 localities
– Average values for: age, household size,
education, car ownership,
– Percentage values for: health, rent,
The best place for students to learn Applied Engineering 9 http://www.insofe.edu.in
SOM on Census Data

CSE 7 30 5 c
Number of data points mapping to each node.
Many blues/reds indicate too large/small model

The best place for students to learn Applied Engineering 10 http://www.insofe.edu.in


SOM on Census Data

CSE 7 30 5 c
Interesting patterns can be seen

The best place for students to learn Applied Engineering 11 http://www.insofe.edu.in


SOM on Census Data

CSE 7 30 5 c
Heat map of average education level

The best place for students to learn Applied Engineering 12 http://www.insofe.edu.in


SOM on Census Data

CSE 7 30 5 c
The best place for students to learn Applied Engineering 13 http://www.insofe.edu.in
SOM on Census Data

CSE 7 30 5 c
Class to take 15 minute break –
run SOM code and suggest other
comparisons

The best place for students to learn Applied Engineering 14 http://www.insofe.edu.in


SOM on Census Data

CSE 7 30 5 c
The best place for students to learn Applied Engineering 15 http://www.insofe.edu.in
SOM on Census Data

CSE 7 30 5 c
Clusters on the map of Dublin

Cluster Nodes using Hierarchical clustering


The best place for students to learn Applied Engineering 17 http://www.insofe.edu.in
References for Kohonen SOMs

• Online tutorials:
– https://www.youtube.com/watch?v=LjJeT7rwvF4
– http://www.cs.bham.ac.uk/~jxb/INC/l16.pdf
– http://www.r-bloggers.com/self-organising-maps-for-customer-se
gmentation-using-r
/
– http://ssdi.di.fct.unl.pt/aadm/aadm1011/slides/LectureSOM.pdf
– http://

CSE 7 30 5 c
www.academia.edu/11322466/Using_R_to_Map_Crime_Density_a
nd_Demographics_in_Boston_from_2012-2014

The best place for students to learn Applied Engineering 18 http://www.insofe.edu.in

You might also like