1
Introduction
1
Perception and Reality
“What is real? How do you
define real? If you’re talking
about what you can feel, what
you can smell, what you can
taste and see, then real is
simply electrical signals
interpreted by your brain. This
is the world that you know.”
—Morpheus’ answer to Neo in
The Matrix, 1999
Compare the ‘Brain in a Vat’
scenario, e.g. Arnold Zuboff’s
‘The Story of a Brain’
1
Some Themes
•
1.
Perception as the construction of a model of the environment
•
2.
Why do we take perception for granted?
•
3. Nature of the model constructed in perception: a symbolic
description (or symbolic representation)
•
4.
•
5. The physiological approach to perception: the perceptual process
as a causal chain
•
6. Parallel approaches to perception from studies of visual
phenomena or visual performance (Psychophysics) and from
physiology and anatomy
•
7. Perception as an Inverse Problem
•
8. Marr’s three levels of Analysis: Hardware, Algorithmic,
Computational…
Perception really is a difficult and impressive accomplishment:
1
Some Themes
1. The world as we perceive it is an internal neural
representation or model, that we construct and rely
on.
1
Some Themes:
2. Why do we take perception for granted?
• If the model is to be useful we have to take it for real:
that’s why we evolved the ability to make it.
• Perceptual ability is universal: not very much
individual variation
• Perception generally requires no conscious effort,
unlike other challenging cognitive accomplishments
1
Some Themes
3.
Nature of the model constructed in perception: a symbolic
description (or symbolic representation)
• A point-to-point projection, preserving the information in the
stimulus, is NOT enough (homunculus fallacy)
• Example of a better approach: ‘bug detectors’
1
Some Themes
4.
Evidence that perception really is a difficult and
impressive accomplishment:
• It takes up nearly half the brain for vision alone
• Infants spend the first year of life mainly learning it
• Engineers can barely achieve even the most primitive
symbolic description with artificial vision systems
1
Early Philosophy of Perception
• Plato’s “The Allegory of the Cave” (380 BCE)
1
Some Themes
5.
The physiological approach to perception: the
perceptual process as a causal chain
•
Phantom limbs
•
Phosphenes
1
Descartes and Phantom Limbs
Descartes:
Reflex arc for action
vs
Consciousness (Pineal )
1
Phosphenes
Tapping in at an intermediate stage of processing:
A visual prosthesis—can use retina, optic nerve or cortex
Compare auditory or visual
‘phosphenes’
1
Some Themes
5. The physiological approach to perception: the perceptual
process as a causal chain
- Phantom limbs
- Phosphenes
6. Parallel approaches to perception from studies of visual
phenomena or visual performance (Psychophysics) and from
physiology and anatomy
– Why these should yield a consistent understanding
7. Perception as an Inverse Problem
- Assumptions/natural constraints: Ames room, etc.
8. Marr’s three levels of Analysis:
- Computational, Algorithmic, Hardware
7. Perception as an Inverse Problem
An example: Use of assumptions/natural constraints (Ames room, etc.)
Without binocular vision, projection still happens…
But now it requires solution of an inverse problem:
What object size and distance produced the given
image? Given one value (image size), we must infer
two.
The mapping from objects to images is many to one.
The restoration of a definite object distance in
perception resolves some of this uncertainty:
-- correctly if the inverse problem is solved correctly
-- incorrectly otherwise
A rational way to attack the inverse problem:
--Choose the most plausible of the alternatives.
--In a Bayesian framework, this would simply be the
most statistically likely, given the context as we know
it. [Although: Hoffman says this is circular logic]
--Example: in Chapter 2, assume uniformity of texture
density in order to compute tilt and slant from the
pattern of non-uniformity in the image
7. Perception as an Inverse Problem
- Use of assumptions about natural constraints:
Ames room(below)
Other constraints: Chapter 2 (texture uniformity constraint), etc…
Are not the Species [images, in modern terminology] of Objects seen with
both Eyes united where the optick Nerves meet before they come into the
Newton, 1682
Brain, the fibres on the right side of both Nerves uniting there, and after
union going thence into the Brain in the Nerve which is on the right side of
the Head [the right optic tract, in present usage], and the fibres on the left
side of both Nerves uniting in the same place, and after union going into the
Brain in the Nerve [optic tract] which is on the left side of the Head, and
these two Nerves meeting in the Brain in such a manner that their fibres make
but one entire Species or Picture, half of which on the right side of the Sensorium
comes from the right side of both Eyes through the right side of both
optick Nerves to the place where the Nerves meet [chiasm], and from thence
on the right side of the Head into the Brain, and the other half on the left
side of the Sensorium comes in like manner from the left side of both Eyes.
For the optick Nerves of such Animals as look the same way with both Eyes
(as of Men, Dogs, Sheep, Oxen, &c.) meet before they come into the Brain, but
the optick Nerves of such Animals as do not look the same way with both
Eyes (as of Fishes and of the Chameleon) do not meet, if I am rightly informed.
Projection (Kepler 1611): 1 world ->2 images ->
1 perceived object (projected into 3D)
Binocular disparity can fix distance:
Descartes 1637 (stick analogy):
Briggs, 1676
Barlow et al., 1968, etc.
Partial decussation:
Newton, 1682 (pub. 1704)
When perceptual constancy is achieved, instead of failing as it does in the Ames
room, the result is a closer correspondence, or isomorphism, between the state
of the external environment and the perceptual impression of it—(closer than the
correspondence between the external environment and the sensory stimulus
through which we know it.)
Action depends on such a simple and orderly mapping, eg x,y vs pixel #;
(Hayek, Kohonen); pointing; Frog computations to control action
--Prediction
--Generalization
--Example of pitch and frequency (and period, and…)
Failure of perceptual isomorphism in intermediate representations:
Perception as a model, with afferent neural representation as data, which needn’t
look like the model, just as a scientific theory should be a model of reality (and
therefore isomorphic with reality in some sense), yet neither the theory nor the
reality it describes has any similarity to the data as such (vectors of numbers,
marks on paper, etc).
A different point of view: Don Hoffman….
8. Marr’s three levels of Analysis:
- Computational, Algorithmic, Hardware
A Hymn to the 3 levels of analysis: “Onward, Marrian soldiers”
Onward, vision scientists,
Marching on with Marr;
Algorithmic theories
Beckon from afar.
Leave aside the neurons
Filling up your head;
Think of computations,
They will do instead.
For those computations,
From their hardware free’d,
Form the firm foundation
Of our noble creed.
Refrain:
Onward, vision scientists,
Marching on with Marr.
Algorithmic theories
Beckon from afar
http://www.hymnsite.com/lyrics/umh575.sht
Text: Sabine Baring-Gould, 1834-1924
Music: Arthur S. Sullivan, 1842-1900
Fear no contradiction
While Marr leads the way;
Views of such abstraction
Few will dare gainsay.
Armed with zero-crossings,
2G our shield,
We can spread confusion
O’er the battlefield.
With its special concepts
Deep and broad and true
Marr’s approach is surely
Good enough for you.
Textbook reference:
Preface, page xi and Ch2, p38