chapter6 - Chapter 6 Self-organizing Map Introduction...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
Chapter 6 Self-organizing Map
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Chapter 6 --- SOM 2 Introduction Introduction Introduction Introduction c Self-organizing feature map : a special class of ANN based on competitive learning , a kind of unsupervised learning very similar to human brain learning process. c The output neurons compete among themselves, only one output neuron (or one neuron per group) is activated at any one time. winner-takes-all competition c Achieve this by lateral inhibitory connections (i.e., negative feedback)
Background image of page 2
Chapter 6 --- SOM 3 Introduction Introduction Introduction Introduction c Neurons are placed at the nodes of a lattice (usually 1-D or 2-D) to ensure that each neuron has a set of neighbours . c Neurons are selectively tuned to various input patterns by competitive learning a meaningful coordinate system for different input features is created over the lattice.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Chapter 6 --- SOM 4 Introduction Introduction Introduction Introduction c The resultant lattice is called the self- organizing feature map or just self- organizing map (SOM). c Similar to a distinct feature of the human brain --- different sensory inputs are represented by topologically ordered computational maps onto different areas of cerebral cortex.
Background image of page 4
Chapter 6 --- SOM 5 Models of SOM Models of SOM Models of SOM Models of SOM c Two different models: Kohonen model c Different from each other in the manner in which the input patterns are specified.
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Chapter 6 --- SOM 6 Willshaw & von der Malsburg Model
Background image of page 6
Chapter 6 --- SOM 7 Willshaw & von der Malsburg Model c 2 separate 2-D lattices of neurons connected together (one projecting onto another). E.g., one is the retina while the other is the visual cortex. c Use short-range excitatory and long-range inhibitory mechanism c Two lattices interconnected by modifiable synapses of a Hebbian type.
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Chapter 6 --- SOM 8 Willshaw & von der Malsburg Model c NOT winner-takes-all: Because a few postsynaptic neurons will fire at any one time. c Weight is limited by an upper boundary condition to prevent network instability.
Background image of page 8
Chapter 6 --- SOM 9 Kohonen Model
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Chapter 6 --- SOM 10 Kohonen Model c More general and commonly-used than c 1-D or 2-D lattice of neurons connected together. The input has connection to each neuron. c Also use short-range excitatory and long-range inhibitory mechanism. c can perform data compression (dimensionality reduction on the input) by optimally places a fixed number of vectors (i.e., codewords) into a higher-dimensional input space.
Background image of page 10
11 Self Self Self-organizing Feature organizing Feature organizing Feature-mapping mapping Algorithm Algorithm Algorithm Algorithm c Goal : transform an incoming signal pattern of arbitrary dimension to a 1-D or 2-D discrete map, and to perform this transformation adaptively in a group of output neurons to be active.
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 12
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 46

chapter6 - Chapter 6 Self-organizing Map Introduction...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online