1. Cogs 160: Neural Coding in Sensory Systems Spring, 2011 Angela Yu http://yulab.ucsd.edu/~cogs160

2. Cell Types in the Retina
Information
Processing

3. Cell Types in the Retina 50+ cell types, distinct in morphology

4. Cell Types in the Retina
Wiring specificity

5. Circuit Diagram of the Retina

6. Classical View of Retinal Ganglion Cells (RGC)
ON-center/OFF-surround
OFF-center/ON-surround

7. Mach Band

8. Detection of Dim Light
Photoreceptors hyperpolarizes, releases inhibition on bipolar cells
Rod bipolar pools over many rod photoreceptors
Band-pass temporal filter minimizes chemical noise
Thresholding (half-wave rectification) minimizes electrical & chemical noise

9. Texture Motion
Y-type ganglion cells sensitive to motion (either direction)
Each shift excites some bipolars and inhibits others
Only active ones contribute, due to thresholding
Biphasic dynamics ⇒ transient response (why?)
Motion prediction: RGC true center (how do we know?)
Continuation when stimulus turns
RGC spatially extended, but no trailing edge (why?)

10. Differential (Object) Motion
Object-motion-sensitive (OMS) ganglion cell is silent under global motion, active under local motion
Built like a Y cell in the center of RF, inhibited by amacrine cell respond to motion in the periphery
OMS activates only when center & surround activities mismatch in time and/or response level

11. Approaching Motion
A type of RGC responds to approaching motion, and not lateral motion
Summing over many ON/OFF bipolar pairs, all thresholded
Leading & trailing edge cancel out in lateral motion
But only leading edge in approaching motion
Gap junction from ON bipolar to Amacrine (extra synpase)

12. Spike Latency Encoding
Spike latency: dark: early, light: late, mixed: intermediate
ON pathway is slower

13. Switching Circuit OFF pathway normally dominant
After saccadic eye movement, ON pathway dominant
Mediated by amacrine cells
Dale’s Law
Why is switching circuit useful?
Decoding?

14. Synaptic Depression & Pattern Adaptation
Synaptic depression due to vesicle depletion (seconds to recover)
B1 & B2 have same RF; A1, A2 inhibit presynaptically
Vertical stripes: B1/A1 asynchronous, strong response, depresses
B2/A2 synchronous, weak response, recovers from depression

15. Linear Readout of Neural Computation
What does a neuron “compute”?
Authors suggest readout should be “easy”?
In particular, linear summation, e.g. summed by a dendritic tree
But dendritic response are known to be nonlinear
And where are x1, x2, ... stored in example?

16. Discussion