Correlated noise and the retina s population code for direction

Transcription

1 Correlated noise and the retina s population code for direction Eric Shea-Brown Joel Zylberberg Jon Cafaro Max Turner Greg Schwartz Fred Rieke University of Washington 1

2 DS cell responses are noisy Stimulus Cell-attached spike rasters ON-OFF DS cells trial 1!! trial 2!!.!.!. Cell 1 3 s

3 DS cell responses are noisy + noise is correlated from cell to cell Stimulus Cell-attached spike rasters trial 1!! trial 2!!.!.!. Cell 1 Cell 2 3 s

4 How does correlated noise impact sensory coding? rate

5 How does correlated noise impact sensory coding? uncorrelated correlated neuron 2 response x x neuron 2 response x x Correlation can degrade signal encoding. neuron 1 response Make sure explain why have elliptical shape -- conc towards diag. neuron 1 response rate Zohary et al 94 Abbott+Dayan 99 Sompolinsky et al. 01, Panzeri + Peterson 01, Averbeck et al., Nat Rev Nsci 06

6 How does correlated noise impact sensory coding? uncorrelated neuron 2 response x x neuron 1 response ν rate Zohary et al 94 Abbott+Dayan 99 Sompolinsky et al. 01, Panzeri + Peterson 01, Averbeck et al., Nat Rev Nsci 06

7 How does correlated noise impact sensory coding? neuron 2 response uncorrelated x x neuron 2 response correlated x x Correlation can enhance signal encoding. neuron 1 response neuron 1 response Varied effects: Depends on signal direction vs. rate νnoise direction. What happens in for DS DS cell cells, circuit, and what are the and circuit why? mechanisms? Zohary et al 94 Abbott+Dayan 99 Sompolinsky et al. 01, Panzeri + Peterson 01, Averbeck et al., Nat Rev Nsci 06 Shamir et al 06 Josic et al 09 Ecker et al 11 da Silvera and Berry, 13 Hu et al 14

8 Population view of DS cell responses - (idealized!) spike count 8

9 Population view of DS cell responses - (idealized!) spike count Signal direction: tangent to sphere Noise direction: Wait to explain idealization until show red arrow. Then, say here s idealization sinusoidal curves w/ same amplitude, offset. 9

10 Population view of DS cell responses - (idealized!) spike count Signal direction: tangent to sphere tangential noise Noise direction: tangential degrades coding 10

11 Population view of DS cell responses - (idealized!) spike count Signal direction: tangent to sphere Noise direction: tangential degrades coding radial enhances coding SAY are two possibilities. Noise could be tangent to sphere, in which case it COULD interfere with signal direction. Or, could be in radial direction, in which case guaranteed to be out of way. tangential noise variance radial variance noise Which case occurs? Measure from cell pairs Model cell population 11

12 Response noise in DS cell pairs Radial Var. Radial Var. + Tangential Var. = Fraction Radial Var. Radial Variance Tangential Variance

13 Response noise in DS cell pairs is largely radial. What are the underlying circuit mechanisms? Radial Var. Radial Var. + Tangential Var. = Fraction Radial Var. Explain scatter plot axes. Then two things to take away. First, points are over 50 percent. Second, correlations contribute strongly to this effect. With correlations Without correlations

14 Paired alternating voltage clamp experiments: conductance correlations inh exc Jon and Fred do intracellular recordings which yield simultaneous measurements of exc and inh, in both cells. That yields key statistics: covariance among all possible conductances within and between cells, and how these depend on stim. This leads to following model cf. Cafaro and Rieke 11

15 Paired alternating voltage clamp experiments: conductance correlations inh mean=gi(s) exc Conductance correlations suggest common input model Common noise source diverges to enter exc and inh pathways for each cell. Along way, multipled by stim-dependent gain. And, that gain is set by the mean response (as for exp nonlin). 1 2

16 Paired alternating voltage clamp experiments: conductance correlations inh exc Conductance correlations suggest common input model 1 2

17 Common input model produces radial variance 1 2

18 Common input model produces radial variance Noise: Nc * g1(s) Nc * g2(s) Nc * g3(s) Mean: g1(s) g2(s) g3(s)... ~n ~m ~m ~n radial variance!

19 Common input model produces radial variance... Data -- show it again Spike outputs preserve radial variance ~m ~n radial variance! model

20 Common input model produces radial variance... Data -- show it again Spike outputs preserve radial variance model data

21 Common input model produces radial variance Bring the sphere back... with pink and blue... argue that the orientation of the cloud matters by rotating

22 Common input model produces radial variance Thus the common input - driven... correlations are responsible for coding being about 180 to 200 pct more accurate r1 r2 r3 that ~doubles coding precision 40 s r5 r6 r4 r3 r2 r1 s est Coding error h s (deg^2) s est 2 i r7 r8 10 Caveat: enough heterogeneity? 0 0 With correlations Without correlations

23 Summary ON-OFF direction selective RGCs give noisy, correlated spike responses Intracellular recordings suggest a common input circuit model This circuit model orients noise in radial direction separates noise from signal, improving coding accuracy

24 Thank you BURROUGHS-WELLCOME FUND SIMONS FOUNDATION NSF -- Math Biol., Robust Intelligence, Statistics Teragrid / XSEDE HHMI Fred Rieke Jon Cafaro Max Turner 71 Joel Zylberberg