1. Biointelligence Laboratory, Seoul National University
Ch 8. Synaptic Plasticity 8.1 ~ 8.4 Adaptive Cooperative Systems, Martin Beckerman, 1997.
Summarized by
Kwonill, Kim
Biointelligence Laboratory, Seoul National University

2. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Contents
8.1 Synaptic Plasticity in the Visual Cortex
8.1.1 Forms of Synaptic Plasticity
8.1.2 Use-Dependent Changes in Synaptic Efficiency
8.2 Models of Synaptic Modification
8.2.1 Feature Selectivity
8.2.2 Stability
8.2.3 Information Processing in the Visual System
8.3 Outline of the Chapter
8.4 Dynamics
8.4.1 Phase Space
8.4.2 Trajectories in Phase Space
8.4.3 Conservative and Dissipative Systems
(C) 2009, SNU Biointelligence Lab, 

3. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
QnA
What is the main model to discuss in this chapter?
(C) 2009, SNU Biointelligence Lab, 

4. Forms of Synaptic Plasticity
Sprouting of new axonal and dendritic structures
Pruning, retraction and disappearance of old arbors
Changes in the membrane properties of existing growth
Cell death
Use-dependent changes in synaptic efficiency
(C) 2009, SNU Biointelligence Lab, 

5. Forms of Synaptic Plasticity – Cell Death
Overproduced neurons  Cell death
15% ~ 85% loss during late prenatal and early postnatal development
1st Function: Numerically matching pre- and postsynaptic neural structures
2nd Function: Selectively removing cells whose axons project to inappropriate targets.
(C) 2009, SNU Biointelligence Lab, 

6. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Forms of Synaptic Plasticity – Use-dependent Changes in Synaptic Efficiency
Critical period in the visual system
Cat, monkey, …
LTP, LTD
Hippocampus of rat
Adult plasticity in which cortical motor and sensory maps
After injury
Not static
STDP (Spike-Timing Dependent Plasticity) ?
(C) 2009, SNU Biointelligence Lab, 

7. Use-dependent Changes in Synaptic Efficiency
Hebb-Stent rule
Highly local process of synaptic modification driven by correlated pre- and postsynaptic activity
Cooperative processes initiated by spatiotemporally correlated afferent input
NMDAR & BCM theory
Critical period
Kitten
(C) 2009, SNU Biointelligence Lab, 

8. Models of Synaptic Modification – Feature Selectivity
Input activity vector:
Random variable
Stationary stochastic process with a time-invariant distribution in general
Synaptic weight vector:
Selectivity of to
averaging
The peak respose to a particular input
(C) 2009, SNU Biointelligence Lab, 

9. Models of Synaptic Modification – Feature Selectivity
Related works
The earliest model by Anderson, Kohonen and Cooper
Linear associator
Bounded weights by Malsburg
Orientation selectivity and related properties
By Perez, Glass, and Shlaer
By Nass and Cooper
(C) 2009, SNU Biointelligence Lab, 

10. Models of Synaptic Modification – Stability
BCM (Bienenstock, Cooper, and Munro)
To explain the development of orientation selectivity and binocular interaction in kitten visual cortex
Single-cell  Network by Cooper and Scofield
Dynamic (nonlinear) equation for synaptic modification by Cooper, Liberman and Oja
Modification threshold (moving)
(C) 2009, SNU Biointelligence Lab, 

11. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Models of Synaptic Modification – Information Processing in the Visual System
Three viewpoints of BCM theory
Single-cell & networks formulation, and its objective function
Connections between the theory and biology
Information-processing functions of BCM
Synaptic systems constraint on the square of the summed synaptic strength
Function as PCA
Maximal information preservation
A number of models of synaptic modification
(C) 2009, SNU Biointelligence Lab, 

12. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Outline of the Chapter
8.4~8.5 : a brief overview of the phase space behavior of dynamic systems
8.6~8.8 : how feature selective states develop from the Hebbian, BCM modification rule, a fixed network architecture, and patterned visual input
8.9~8.10 : the neurophysiological basis for the theory
8.11~8.13 : the information processing activities carried out by model neurons in the visual system
(C) 2009, SNU Biointelligence Lab, 

13. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Dynamics – Phase Space
Phase space
a space in which all possible states of a system are represented, with each possible state of the system corresponding to one unique point in the phase space
In BCM theory, plane is interesting
An physical example : Newton’s 2nd law
(C) 2009, SNU Biointelligence Lab, 

14. Trajectories in Phase Space
Fixed points
Equilibrium states of thermodynamic system
Limit cycle
Periodic motion (oscillator)
(C) 2009, SNU Biointelligence Lab, 

15. Conservative and Dissipative Systems
Properties of phase space
Trajectories in phase space are unique
No intersection
Collections of neighboring points obey certain invariance conditions
A theorem of Cauchy’s
Liouville’s theorem
(C) 2009, SNU Biointelligence Lab, 

16. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Summary
Synaptic Plasticity
Cell Death
Use-Dependent Changes in Synaptic Efficiency
Models of Synaptic Modification
Feature Selectivity
Stability : BCM theory
Information Processing in the Visual System
Outline of the Chapter
Dynamics
Phase Space
Trajectories in Phase Space
(C) 2009, SNU Biointelligence Lab, 

17. (C) 2009, SNU Biointelligence Lab, http://bi.snu.ac.kr/
QnA
What is the main model to discuss in this chapter?
BCM theory
(C) 2009, SNU Biointelligence Lab,