V1 Physiology

1
V1 Physiology
2
Questions
Hierarchies of RFs and visual areasIs
prediction equal to understanding? Is predicting
the mean responses enough?General versus
structural models? What should a theory of V1
look like? How is information represented in V1?
3
The cortex
4
Visual Areas in the Nonhuman Primate
Felleman van Essen
5
Visual Areas in the Nonhuman Primate
6
Monkey LGN
7
Monkey LGN
8
Monkey V1 Laminar organization
9
Monkey V1 Inputs
10
Monkey V1 Outputs
11
Monkey V1 Oculodominance Columns
12
Monkey V1 Oculodominance Columns
13
Monkey V1 CO patches (or blobs)
14
Monkey V1 Orientation Tuning
15
Monkey V1 Orientation Columns
16
Monkey V1 Orientation Map
What generates the map? How does it develop?
What is the role of experience? What is its
functional significance (if any)? How are
receptive field properties distributed with
respect to the map features (such as
pinwheels)? What is the relationship to other
maps (retinotopy)?
17
Monkey V1 The Ice Cube Model
18
LGN cell
19
V1 simple cell
20
V1 complex cell
21
Hierarchy of Receptive Fields
22
Simple cells receptive fields
23
Models v0.0
24
Monosynaptic connectivity from thalamus to
layer 4
Alonso, Usrey Reid (2001)
25
Monosynaptic connectivity from thalamus to
layer 4
Reid Alonso (1995)
Alonso, Usrey Reid (2001)
26
Expected response of linear RF to moving
gratings
27
Yet F1/F0 distributions are bimodal
Skottun et al (1991)
28
There appears to be a continuum of responses
Priebe et al, 2004
29
Beware of bounded indices
Priebe et al, 2004
30
Laminar distribution of F1/F0
Same in cat (Peterson Freeman but see Martinez
et al)
31
Standard Models v1.0
32
Stochastic stimuli
Conditional Stimulus Distributions
P(s)
P(s spike)
How are the original and conditional stimulus
distributions different?
33
Standard Models v1.1
34
Elaborating the LN model
35
Simple-cell nonlinearities Saturation
Carandini, Heeger Movshon (1996)
36
Saturation depends on orientation
Carandini, Heeger Movshon (1996)
37
Simple-cell nonlinearities Masking
Carandini, Heeger Movshon (1996)
38
Non-specific gain control can shape tuning
selectivity
39
Prediction Understanding?
40
The linear-nonlinear model
41
Simple cell receptive fields in V1
42
Simple cell receptive fields in V1
43
Simple cell receptive fields in V1
44
Simple cell receptive fields in V1
45
Simple cell receptive fields in V1
46
Why this particular set of filters?
47
Going beyond the modeling of mean responses
Cortical State,
Stimulus,

• The response to sensory stimulation at any one
  time is a function of both the recent history of
  the stimulus and the cortical state.
• If the ongoing cortical activity is noise then
• Measure the mean response to sensory stimulus
• Measure how the mean response varies with
  stimulus parameters.

48
The vending machine analogy
Current State,
Stimulus,
Count up to 75 and deliver a coke (a
deterministic machine)
49
The vending machine analogy
50
25
0
Count up to 75 and deliver a coke (a
deterministic machine)
50
The vending machine analogy
Current State,
Stimulus,
Count up to 75 and deliver a coke (a
deterministic machine)
51
The vending machine analogy
52
The vending machine analogy
53
Modeling the Mean Response Is it sufficient?
Arieli et al (1996)
Mean response
Single trial prediction
Single trial response
54
Modeling the Mean Response Is it sufficient?
Supèr et al (2003)
55
Seeking invariants of the population response
Stimulus
Response
Percept
Vertical grating
Vertical grating
Vertical grating
There must be some invariant feature in the
population responses. Asking about the neural
code is equivalent to asking what is this
invariant (best clustering approach of Victor
et al).
56
Theory of Visual Area X
Representation Area X is about representing
natural signals optimally. Estimation/Bayes Area
X is all about estimating the most likely
stimulus (motion/contours/etc) given the
statistics of natural signals. Processing Area X
is doing some interesting image processing (for
example, face detection) Behavior Area X is
about using visual information for visually
guided behavior (active vision)
57
Half-Time