Table 3 Prototype synthesis experiments.
From: Visual prototypes in the ventral stream are attuned to complexity and gaze behavior
V1 | V4 | IT | |
|---|---|---|---|
Monkey A | |||
Median change in firing rate to adaptive synthetic images (no. experiments) | 36.5 ± 9.0 (29) | 24.8 ± 7.5 (28) | 16.2 ± 4.1 (70) |
Median change in firing rate to fixed reference images | −2.0 ± 2.1 | −3.1 ± 1.6 | −2.3 ± 0.6 |
Percent of successful experiments, defined as those where the final synthesized images evoked higher responses than the first-generation images (95% CI of difference not including zero) | 96.6 | 85.7 | 71.4 |
Percent of successful experiments, defined as those where the final synthesized images evoked higher responses than the reference images (Wilcoxon rank-sum test, one-sided, P < 0.02) | 96.6 | 78.6 | 61.4 |
No. generations needed to reach half-maximum mean response (Wilcoxon rank-sum test, two-sided, contrasting V1/V2 and IT) | 10.5 ± 1.0 | 13.7 ± 1.1 | 17.5 ± 2.2 P = 1.6 × 10−33, Z-value 12.1, N = 99 experiments |
Monkey B | |||
Median change in firing rate to adaptive synthetic images (N experiments) | 54.9 ± 15.7 (35) | 46.4 ± 6.3 (30) | 50.5 ± 11.6 (60) |
Median change in firing rate to fixed reference images | −5.9 ± 1.6 | −4.8 ± 4.1 | −6.7 ± 1.3 |
Percent of successful experiments, defined as those where the final synthesized images evoked higher responses than the first-generation images (95% CI of difference not including zero) | 91.4 | 93.3 | 83.3 |
Percent of successful experiments, defined as those where the final synthesized images evoked higher responses than the reference images (Wilcoxon rank-sum test, one-sided, P < 0.02) | 88.6 | 66.7 | 55.0 |
No. generations needed to reach half-maximum mean response (Wilcoxon rank-sum test, two-sided, contrasting V1/V2 and IT) | 12.0 ± 2.6 | 14.6 ± 2.7 | 16.4 ± 1.7 |
P = 4.8 × 10−33, Z-value 12.0, N = 95 experiments | |||
