1
(No Transcript)
2
Memory is deceptive because it is colored by
today's events. (Albert Einstein)

• David Hadas, CS TAU
• Under the supervision of
• Prof. Nathan Intrator, Dr. Galit Yovel

3
Agenda

• Distinction
• The magical number 7
• Multidimensional stimulus identification
• Recognition
• RBC
• Memory
• Effects of temporal association on recognition
  memory
• Morphing visual memories through gradual
  associations
• Our research
• Working memory
• Long term memory
• Identifying object classes
• Discrete vs. Continuous memory
• Current work

4
The magical number seven George A.
Miller (1956)

• For unidimensional judgments The span of
  absolute judgment is 7
• We can improve by
• Making relative judgments
• Increasing dimensionality
• Using successive judgments

3 bits
2.5 bits
Communication Channel
In
Out
Source Pollack 1952, 1953
5
Multidimensional stimulus identification
Egeth Pachella 1969

• Slower
• More error prone
• Speed and Accuracy decline with dimensionality

6
Recognition By Components (RBC) Irving
Biederman (1987)

• Primal Access
• Memory Access
• The first contact of perceptual input from an
  isolated unanticipated object to a representation
  in memory

Edge Extraction
7
Recognition By Components (Cont)

• 36 Geons suggested

8
Effects of temporal association on recognition
memory G. Wallis, HH Bulthoff 2001

• We are continuously associating views of objects
  to support later recognition.

9
Morphing visual memories through gradual
associations S. Preminger, D. Sagi, M Tsodyks
(unpublished)

• How exposure to face stimuli, associated with a
  previously memorized face , influences the
  long-term memory of the stored face

10
Our Research
11
Storing a representation

• We cannot memorize what we cannot perceive
• We can only store what we have in our working
  memory

Representation in Long Term Memory
Store
12
Recalling a representation

• We cannot recall what we cannot perceive
• We can only recall what we have in our working
  memory
• Do we store as we retrieve?

13
Working memory

• We have limited ability to measure
• Can we at least maintain a measurement?

14
Test your working memory

• Find the longest line

The Test Ended
Jump to next test
15
Study the size of this circle (1)
Jump to next test
16
How many circles of the same size do you see? (1)
The Test Ended
Jump to next test
17
Results (1)
Jump to next test
18
Study the size of this circle (2)
Jump to next test
19
How many circles of the same size do you see? (2)
The Test Ended
Jump to next test
20
Results (2)
Jump to next test
21
What is the difference between the tests?

• In the second test
• All circles are small
• We modify the internal representation based on
  the stimuli

Jump to next test
22
A representation in working memory

• Depending on the memory type
• We have limited ability to reliably store even a
  single memory
• We change the representation based on the stimuli

23
Long term memory

• We seem to have a slight working memory problem
• We cannot reliably remember even a single
  dimension
• Can we at least trust our long term memory to
  help out?

24
How detailed is our long term memory?

• Can you describe a table?
• Can you describe your table at home?
• How do you remember it?
• Can you describe Clinton?
• How do you remember him?

25
We remember discretely

• Discrete
• Parallel or not
• Symmetrical or not
• Collinear or not
• Curvlinear or not
• Type of vertex
• Number of edges
• Number of vertices
• Certain Components
• Certain component arrangements

• Continuous
• Length/Width/Depth
• Angle
• Ratio
• Color

• Semi-Discrete
• 7 Categories (?)
• Categories of
• Length/Width/Depth
• Angle
• Ratio
• Color

26
A representation in long term memory

• We can only store what we have in our working
  memory
• We change the representation based on the stimuli
• We seem to store as we search our long term
  memory
• As a result
• Our long term memory is adaptive
• We cannot rely on it for continuous measurements

27
Current work
Protocols and measurements for tuning of the
brain neural network

• We develop a tool and a technique for qualifying
  and quantifying the plasticity shown during
  stimuli morphing
• We study
• The brain inability to secure preset object
  representations
• The brain natural and involuntary tendency to
  modify such representations based on stimuli
• The affect of a protocol on the internal
  representation
• We measure plasticity of subjects exposed to a
  morphing protocol
• We characterize the affecting protocol and the
  resulting changes

28
Could it be thatEinstein was right after
all?When he said
Memory is deceptive because it is colored by
today's events. (Albert Einstein)
EMC2