Fig. 5: Input variance modulates orientation decoding in V1.
From: Cortical recurrence supports resilience to sensory variance in the primary visual cortex
a Time course of orientation θ decoding accuracy at two variances Bθ. Lines are the mean accuracy of n = 5 random resampling of 100 neurons and contour the SD. Significantly better decoding from resilient neurons at Bθ = 35° is shown as a gray overlay (Wilcoxon signed-rank test, threshold p < 0.01). Decoding at chance level is represented by a gray dashed line and stimulation time by a black line. b Population tuning curves with a von Mises fit, showing the likelihood of decoding each θ in four time windows. c Same as (a) for the two groups of neurons. d Same as (b) for the two groups of neurons, with Bθ = 0° (upper row) and Bθ = 35° (lower row). e Time course parameters for three decoders at all Bθ, estimated by fitting a sigmoid up to PST = 300 ms. τ is the time constant. f Correlation between classification accuracy and population circular variance for the whole population (left), for both groups with Bθ = 0° (middle) and Bθ = 35° (right). Linear regression is shown as solid lines with slope m indicated (all significant, p < 0.001, Wald Test with t-distribution).
