Testing and Re ning a Computational Model of Neural Responses in - - PowerPoint PPT Presentation

▶
testing and re ning a computational model of neural
SMART_READER_LITE
LIVE PREVIEW

Testing and Re ning a Computational Model of Neural Responses in - - PowerPoint PPT Presentation

Feb 01, 2023 117 likes •282 views

Testing and Re ning a Computational Model of Neural Responses in Area MT Eero Simoncelli Wyeth Bair James Cavanaugh Tony Movshon Computer & Information Science Dept, U. Pennsylvania Howard Hughes

slide-1
SLIDE 1

Testing and ReÞning a Computational Model

  • f Neural Responses in Area MT

Eero Simoncelli

  • Wyeth Bair
  • James Cavanaugh
  • Tony Movshon
  • Computer & Information Science Dept, U. Pennsylvania
Howard Hughes Medical Institute, NYU Center for Neural Science, NYU
slide-2
SLIDE 2

Stage I: V1 Simple Cells

+

  • . . .

. . . . . .

e-2 Input: image intensities

. . .

Combines image intensities via space-time oriented linear Þlters. Response tuned for spatio-temporal frequency.

5/96 1

slide-3
SLIDE 3

MT Component Cell Construction

  • + + +
  • + + +
  • + + +
  • Linear Combination of V1 Cells tuned for:

– range of spatial frequencies / R.F. positions – Þxed orientation and speed

RectiÞcation / Normalization
slide-4
SLIDE 4

MT Component Cells

Input: stage I outputs c-2

  • . . .

. . . . . . +

c

  • . . .
Combines outputs of V1 units tuned for different spatial and tem-

poral frequencies, and RF positions.

Response tuned for orientation and speed of 1D patterns (stripes).

5/96 2

slide-5
SLIDE 5

MT Pattern Cell Construction

  • + + +
  • + + +
  • + + +
  • +

+ +

  • + + +
  • +

+ +

  • s

s s/2 s/2 s/2 s/2

Linear Combination of V1 Cells tuned for:

– range of spatial frequencies / R.F. positions – speed/orientationcombinationsconsistentwith pattern motion

RectiÞcation / Normalization
slide-6
SLIDE 6

MT Pattern Cells

Input: V1 outputs v-2

  • . . .

. . . . . . +

v

  • . . .
Combines outputs of V1 units tuned for different orientations, spa-

tial/temporal frequencies, and RF positions.

Response tuned for 2D image pattern velocity (speed & direction).

5/96 3

slide-7
SLIDE 7

Component vs. Pattern

  • 180

180

  • 180

180

Component Cell Grating Plaid

  • 180

180

  • 180

180

Pattern Cell

Movshon et. al., 1986

5/96 4

slide-8
SLIDE 8

Drifting Dot Stimulus

Sum of Gratings

=

s s √3 s/2 √3 s/2 s/2 s/2

+

5/96 5

slide-9
SLIDE 9

Drifting Dot Stimulus

Sum of Gratings

direction

t = s * s t

  • 90
  • 60
  • 30

30 60 90

5/96 6

slide-10
SLIDE 10

Component Cell / Drifting Dots

direction speed slow dots medium dots fast dots direction direction cell preference stimulus

Prediction: direction-tuning for dots becomes bimodal at high speeds.

5/96 7

slide-11
SLIDE 11

Component Cell / Drifting Dots

Cell Model

  • 180
  • 120
  • 60

60 120 180

direction speed

6.5 d/sec 3.25 d/sec 1.63 d/sec 0.81 d/sec 0.4 d/sec

  • 180
  • 120
  • 60

60 120 180

direction

Bimodality at high speeds (as predicted). Similar behavior in V1 complex cells (Movshon,
  • et. al., 1980).
slide-12
SLIDE 12

Component Cell / Sine Grating

Cell Model

0.21 Hz 0.41 Hz 0.83 Hz 1.65 Hz 3.3 Hz

direction

  • 180
  • 120
  • 60

60 120 180

t

  • 180
  • 120
  • 60

60 120 180

direction

Direction-tuning curves independent of speed.
slide-13
SLIDE 13

Pattern Cell / Drifting Dots

Cell Model

6.5 d/sec 3.25 d/sec 1.63 d/sec 0.81 d/sec

  • 180
  • 120
  • 60

60 120 180

0.4 d/sec

direction speed

  • 180
  • 120
  • 60

60 120 180

direction

Direction-tuning curves independent of speed.
slide-14
SLIDE 14

Pattern Cell / Sine Grating

Cell Model

  • 180
  • 120
  • 60

60 120 180

0.21 Hz 0.41 Hz 0.83 Hz 1.65 Hz 6.6 Hz

direction

t

  • 180
  • 120
  • 60

60 120 180

direction

Bimodality at low speeds (as predicted). Bimodalitynotfoundforsingledriftingbars(Rod-

man & Albright, 1987).

slide-15
SLIDE 15

Conclusions

Stimulus

n Cell Type

Component Pattern Grating Unimodal Bimodal @ low speeds Dots Bimodal @ high speeds Unimodal

5/96 8