Scribd
Upload
26 views19 pages

Lecture 11 Unsupervised Learning

This document provides an introduction to unsupervised learning algorithms. It begins by explaining that unsupervised learning is used when target values are difficult to obtain from training data. The goal is to find clusters of similar inputs. Examples given include identifying weather patterns and protein structures. Main algorithms covered are clustering methods like K-means, K-medoids, SOM (Self Organizing Maps). SOM is described in more detail, explaining how it is inspired by brain processing and can maintain topology while training via competition between neurons. Applications of SOM include dimension reduction and detecting similarity. The document concludes with a group activity to explain other clustering techniques.

Uploaded by

Hodatama Karanna One
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
26 views19 pages

Lecture 11 Unsupervised Learning

This document provides an introduction to unsupervised learning algorithms. It begins by explaining that unsupervised learning is used when target values are difficult to obtain from training data. The goal is to find clusters of similar inputs. Examples given include identifying weather patterns and protein structures. Main algorithms covered are clustering methods like K-means, K-medoids, SOM (Self Organizing Maps). SOM is described in more detail, explaining how it is inspired by brain processing and can maintain topology while training via competition between neurons. Applications of SOM include dimension reduction and detecting similarity. The document concludes with a group activity to explain other clustering techniques.

Uploaded by

Hodatama Karanna One
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/ 19

11

CS7113
Machine Learning
W. M . Kalani Sriya Ilmini
Department of Computer Science
Faculty of Computing
Sir John Kotalawela Defence University
Rathmalana

1
Unsupervised Learning
Lecture 11

2
Introduction
• The learning algorithms we learnt were supervised learning
algorithms.
• We used a training data set that consists of a collection of labelled target
data.
• But in many situations, it is difficult to obtain the target values.
• In such situations, we need to identify similarities between inputs
that belong to the same class (initially it is not known).
• The aim of the unsupervised learning algorithms is to find clusters of
similar inputs automatically.

3
Examples
• E.g. 1:
• What weather patterns drive certain ecological process?
• What is the underlying amino acid sequence that produced the
specific structure of a protein?
• What types of species insects, fish, plants, and so on assemble
together, and what are the conditions under which they form
assemblages?

4
Algorithms
• Clustering
• K-Means Algorithm
• K-Medoid Algorithm
• K-Means Neural Networks
• SOM

5
SOM: Self Organizing Maps
• Data clustering technique
• Dimension reduction technique
• Competitive learning technique
• the neurons in the SOM learns by competing with each other to become the
winner

6
SOM…
• SOM is inspired by how our brain process sensory data.
• Brain cortex has divided itself to separate areas to process sensory
data (visual data are processed in visual cortex, acoustic are
processed in auditory cortex).
• This means for the same input, same area of the brain is activated.
Different signals activate different parts.
• In SOM we design a neural network which has the ability to activate
similar areas for similar inputs and different areas for different inputs

7
SOM…
• There are two novel departures from the traditional neural networks:
1. The relative locations of the neurons in the network matters- this property
is known as feature mapping- nearby neurons correspond to similar input
patterns.
2. The neurons are arranged in a grid (usually 2D, but sometimes 1D line is
also used) with connection between the different layers.

8
SOM…
• Maintains the topology of the dataset
• Training occurs via competition between the neurons
• Impossible to assign network nodes to specific input classes in advance
• Can be used for detecting similarity and degrees of similarity
• Random weight vector initialization

9
Components of the SOM
• Sample data

• Weights

• Output nodes
10
Structure of the map
• 2-dimensional or 1-dimensional grid

• Each grid point represents an output node

11
Structure of the map
• Initialize Map
• For t from 0 to 1
• Select a sample
• Get best matching unit
• Scale neighbors
• Increase t a small amount
End for

12
Determining the BMU Neighborhood

13
SOM- Application

14
SOM- Application…

15
SOM- Application…

16
SOM- Application…

17
Unsupervised Learning – Group work
• Explain the other clustering techniques which were mentioned at the
start of the lecture.
• What is it?
• How they work?
• If any pros and cons

• Each group has to present their ideas within 5 mints.

18
Thank You!

You might also like