Read CCN_adaptation text version

A- 'Unaware' When neurons and perception change (2) A- 'Unaware'

Adaptation! Adaptation Adaptation! State = history of State stimulation

Adaptation

s s

Further reading: A. Kohn, Visual adaptation: physiology, mechanisms, and functional benefits, J Neurophysiol, 2007.

Population ! Population ! Response Response

!! !!

Encoder Encoder

A ! Tuning Curves

r r

Fixed ! Fixed ! ! Decoder ! Decoder

s ^ s ^

B ! Population Response 50

B- 'Aware' B- 'Aware'

25 0 !180 !90 0 !test 90 180

Responses

50

25

0 !180 !90 0 !test 90 180

Adaptation! Adaptation! State State Population ! Population ! Response Response

!! !!

s s

Motion After-Effect

The Waterfall Illusion

known to the Ancient Greeks, but the first modern report of it is often attributed to Robert Addams (1834), who observed the effect while viewing a waterfall at Foyers in Scotland.

Encoder Encoder

r r

Adaptive ! ! Adaptive ! Decoder ! Decoder

s ^ s ^

Orientation After-Effect

Look at the circle for 30 sec

Then this one

similarly, http://www.michaelbach.de/ot/mot_adapt/index.html

The "Tilt after-effect"

ally make when viewing faces. e classifying a face image accorddual identity, and may involve ng the stimulus continuum in ies) for such characteristics11­13. r categorical perception of faces, mulus boundaries that observers ories can be shifted by prior

ng between pairs of face images to mages spanning the two original gender, ethnicity (Japanese or . Observers made forced-choice ng the continuum (for example, from a gender morph appeared nses were used to estimate the response was equally likely. We s shifted after observers viewed defining the categories. resents an androgynous image ale exemplars, and could be set to the magnitude of the shifts after adapting to a male face, the ared distinctly female, and thus ral was shifted towards the male ation to the female face induced in the gender boundaries after ndividual subject with a two-way e (male versus female) by test five observers, the mean settings emale adaptation (F 1,18 $ 23.36, or the specific face pair tested ) and did not show a significant ition and face pair (F 2,18 # 1.08, cts for each of the other face ple, adaptation to a Japanese or y biased the perceived neutral them (F 1,18 $ 56.1, P , 0.001) e perceptually (Fig. 3), and are eived neutral point towards the apting face itself appears more may re-centre the `face space' figuration. mulus differences defining traits male face pairs or only to the ntity? To assess this we adapted male or ten female faces, each

original expression. The after-effects of adaptation were consistent with those we found for gender and ethnicity (Fig. 2d). For example, in a happy­angry morph, previous adaptation to the angry or happy face induced clear differences in the expression boundary (F 1,6 $ 9.71, P , 0.05). Similar shifts in the neutral point after adaptation occurred in morphs between disgust­surprise (F 1,6 $ 18.54, P # 0.01) or fear­contempt (F 1,6 $ 11.78; P , 0.05). In the course of these measurements we also discovered large individual differences in the category boundaries chosen during the pre-adaptation settings, and these appeared to be related to the categories to which the observers themselves belonged. This is

a psychophysical procedure of temporal two-alternative forced choice (2AFC). Alternatively, the orientations of two con-

tours may be compared during a steady fixation, so that the

0)

>

two orientations stimulate different retinal regions at the same

I-

z

co,

time. This procedure corresponds fairly closely to a psychophysical procedure of spatial 2AFC. It has yet to be established whether these two procedures give substantially different estimates of orientation discrimination. The two procedures may even involve different discrimination prostimulate the same retinal Visual Adaptation: compared tothat compares two spatially region andtargets. In Psychophysics separated the other process one

10 20 30 cesses, one a local process that requires the patterns to be

Face after-effects

!

VisualANTICLOCKWISE adaptation leads0 to: -30 -20 -10 VERTICAL After viewing male faces, subsequent faces look more female. After viewing caucasian faces, subsequent faces look more asian. thin/fat etc..

addition, when the two gratings are presented to the same location in quick succession, the temporal 2AFC procedure

used in most studies of orientation discrimination can include

Fig. 2. Acuity of orientation discrimination for different azimuths. Ordinates plot reciprocal of discrimination threshold. Continuous line and dashed line plot preadaptation and postadaptation data, respectively. Each point represents a 35-min run.

1001

a stimulus motion, whereas the spatial detection tasks: higher detection thresholds for apparent rotary there is strong evidence that TEST GRATING ORIENTATION deg. 2AFC procedure does not, and

the visual pathway contains mechanisms specifically sensitive ple, therefore, the characteristics of orientation discrimination in everyday vision may depend on whether successive fixation

2 2 to motion 3 and possibly even to rotary motion. 4 In princi-

CLOCKWISE

0 z I0

UJ

or steady fixation is used (and, in laboratory studies, whether temporal or spatial 2AFC is used). This study is restricted to temporal 2AFC. 150

0

75

0 -J

0

a:

I

z

r

50

100 W -J

w

0 -J

0

25

50 aI Co

The more similar the test neurally of the following four physiological operations: (1)to the encode the orientation the first grating,the increase adaptor, of the higher (2)store the neural representation of orientation, (3) neurally encode the orienin detection threshold tation of the second grating, (4) compare the neural representations of the two orientations. One remarkable aspect of orientation discrimination, its

high acuity, is well known. No less remarkable, though, is that

One way of describing orientation discrimination is in terms

z5

0 a:

0

0

I

a, cc

z

0

Figure 1 Examples of face pairs for the dimensions of gender (top row), ethnicity (middle row), or expression (bottom row). Morph arrays consisted of 100 images spanning the two original faces, which were assigned a value of 0 or 100. The morph level (for example, 25, 50, or 75) therefore corresponded to how far along the sequence the image fell between the two originals. Category boundaries corresponded to the image level chosen as the neutral point within the morph sequence.

NATURE | VOL 428 | 1 APRIL 2004 | www.nature.com/nature

z 0 0

discrimination threshold changes only slightly or not at all when the interval between the two grating presentations is increased from 1 to 10 sec. This implies not only that the neural representation of the first grating's orientation is sufficiently precise to sustain a 0.15-0.5-deg discrimination threshold but also that this neural representation is stored with little decay in its precision over time to at least 10 sec.

Orientation Discrimination and Contrast Detection: Opponent-Process and Line-Element Models of Discrimination

z -25 w

-30 -20 -10 0 10 20 30

CLOCKWISE

ANTICLOCKWISE

VERTICAL

©2004 Nature Publishing Group

Webster, Nature, 2004

5

TEST GRATING ORIENTATION deg. Fig. 3. Postadaptation threshold elevations for orientation discrimination (continuous line) and for contrast detection (dotted line). The adapting grating was vertical (0 on abscissa).

Our main findings are that (1) adapting to a high-contrast sine-wave grating improves orientation discrimination at an orientation parallel to the adapting grating while simultaneously degrading contrast detection; (2) adaptation elevates

discrimination thresholds at angles of about 11-17 deg to ei-

[Regan and Beverley, 1985]

Visual Adaptation: Psychophysics

Visual adaptation leads to: ! detection tasks: higher detection thresholds ! estimation tasks : strong biases (mainly repulsion)

Visual Adaptation: Psychophysics

Visual adaptation leads to: ! detection tasks: higher detection thresholds ! estimation tasks : strong biases (mainly repulsion)

Repulsion

40 30 20 Bias (°) 10 0 !10 !20 !30 !180 !90 0 90 Angle from Adaptor (°) 180

Patterson 96 Levinson 76 Alais 99

2.5

Phinney 97 Hol 01

Threshold ratio

2 1.5 1 0.5 !60

!30 0 30 Angle from Adaptor (°)

60

Visual Adaptation: Psychophysics

Visual adaptation leads to: detection tasks: higher detection thresholds estimation tasks : strong biases (mainly repulsion) ! discrimination tasks: changes in discrimination thresholds.

! !

High Contrast Adaptation

Adaptation to 100% contrast : reduces apparent contrast for all test contrasts ! changes discrimination threshold

!

50

5

Langley 02 Ross 91

30 20 Bias (°) 10 0 !10 !20 !30 !180

Threshold ratio

Patterson 96 Levinson 76 Alais 99

Thres. Ratio

Bias (%)

40

2.5 2 1.5 1 0.5 !60

Impaired discrimination Phinney 97 at flanks

Hol 01

25

4 3 2 1

0

Impaired discrimination at low contrast

Greenlee 88 Abbonizio 02

No change or Slightly Better discrimination

!25

Modest improvement at high contrast

20 40 60 Contrast (%) 80 100

!50

20

test contrast (%)

40 60 Contrast (%)

80

100

test contrast (%)

!90 0 90 ° Angle from Adaptor ( )

180

!30 0 30 ° Angle from Adaptor ( )

60

Bias

20 1.5 1 0.5 1.5 1 0.5

10 Thres. Ratio

A

0 -20

Visual Adaptation: Physiology

!

20

B

0 -20 5 1.5 0 1 -5 0.5 5 1.5 0 1 -5 0.5 -180

Neural basis of Contrast adaptation

For motion direction, orientation, ... (bell-shaped tuning curves)

A ! Tuning Curves

Bias B ! Population Response

20

Thres. Ratio

4

Responses

50

50 A

C

!

0 2 0 1.5

For sigmoidal response curve: Mainly a shift towards the adaptor

-20

25

25

B

20

D

A

1 0.5 -180

B 100

0

100 control adapt post!adapt ! (%) 10 Contrast (%) 100 Mean Response

-90 0 90 180

0 !180

0 !90 0 !test 90 180 !180

C

-20

-90

0

90

180

75

!90

i - adapt

Mainly a Gain change

[Van Wezel & Britten 2002, Krekelberg et al. 2006]

0 5 !test

0

90

180

4 2 0 1.5 1 0.5 -180

D

Other effects are controversial, -5 dependent on time scale and area: shifts in preferred5 orientation, changes in width,0 changes in variability. -5 -180 -90 0 90 180 [Kohn & Movshon 2004, Dragoi et al, 2000]

50

50

25

0 1

-90 0 90 180

0 0

50 pre!adapt ! (%)

100

12

Mechanisms

Question: Cellular mechanism ? Network effect? Still debated. - Intracellular recordings in cat V1 show that contrast adaptation leads to a large hyperpolarization of the membrane potential (Carandini & Ferster, 1997), which is at least partly due to cellular mechanisms (activation of sodium-gated potassium channels, Sanchez-Vives et al 2000). - Synaptic depression (Chung et al 2002), due to depletion of vesicles from presynaptic terminal. Thalamo-cortical synapses only? corticocortical?

Temporal scales

· Duration matters. · Some effects appear after very short durations, e.g. 300 mec,

specially at higher processing stages. · To a first approximation, adaptation effects appear qualitatively similar on a wide range of time scales with more prolonged exposure resulting in stronger effects [Kohn, 2007]

13

14

Functional role ?

! !

Conclusion

! Adaptation

reduce metabolic costs.

is a form of plasticity, common feature of cortical causes significant changes in perception

responses, on multiple time scales.

! Adaptation ! Adaptation

improves coding (discriminability) of most frequent conditions? = re-center tuning around prevailing stimulus conditions? ! luminance adaption leads to increase in discrimination to match prevailing conditions, perceptual benefits of other types of adaptation are unclear (weak enhancements in discriminability) improves detectability or discriminability of novel or rare stimuli? ! weak evidence for improvement of detectability or discriminability for novel stimuli.

!

is used as a tool to study underlying representations (`the psychologist's microelectrode'), e,g. nowadays, specificity and invariance in fMRI.

! !

biophysical mechanisms are still murky locus of adaptation often unclear.

! An

appealing hypothesis is that adaptation serves efficient coding, to match the response properties of our sensory systems to prevailing environment.

! A better

understanding of the relationship between neural

15

responses and perception will help validate this assumption.

Information

CCN_adaptation

4 pages

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

1271459


You might also be interested in

BETA
CCN_adaptation