Contextual associations facilitate object recognition in human vision. However, the role of context in artificial vision remains elusive as does the characteristics that humans use to define context. We investigated whether contextually related objects (bicycle-helmet) are represented more similarly in convolutional neural networks (CNNs) used for image understanding than unrelated objects (bicycle-fork). Stimuli were of objects against a white background and consisted of a diverse set of contexts (N=73). CNN representations of contextually related objects were more similar to one another than to unrelated objects across all CNN layers. Critically, the similarity found in CNNs correlated with human behavior across three experiments assessing contextual relatedness, emerging significant only in the later layers. The results demonstrate that context is inherently represented in CNNs as a result of object recognition training, and that the representation in the later layers of the network tap into the contextual regularities that predict human behavior.

Background and Objective Hypothesizing that stress dysregulation may worsen cocaine dependence, we investigated the effect of diurnal cortisol secretion profile, suppression of cortisol secretion, and total cortisol secretion on retention, abstinence-based voucher earnings, days of cravings, and mood status of participants at the end of a 2-week medication-free lead-in prior to randomization in a clinical trial of mirtazapine (60 mg vs. placebo) for depressed cocaine-dependent patients. Methods We measured saliva cortisol levels at 9am, 2pm, and 5pm on the first two consecutive days of a 2-week medication-free lead-in period. Results from saliva samples were used to estimate the total daily level of cortisol, the diurnal profile of secretion (typical vs. atypical), and response to dexamethasone suppression (0.1 mg). Seventy-seven patients collected saliva samples at baseline, and 65 (85%) were suitable for profile analysis. Results Patients with typical profiles (52%) collected significantly more abstinence-based voucher earnings during the lead-in (U = 299.50, p = .025). Diurnal secretion profile did not significantly affect mood status, days of craving, or retention. There were no significant effects of suppression of cortisol secretion or of total cortisol levels on any outcome measures. Conclusion In a subgroup of cocaine-dependent patients, deviation of cortisol secretion away from the homeostatic diurnal pattern was associated with reduced success at achieving early abstinence, an important determinant of treatment success.

No abstract

Contextual associations facilitate object recognition in human vision. However, the role of context in artificial vision remains elusive as does the characteristics that humans use to define context. We investigated whether contextually related objects (bicycle-helmet) are represented more similarly in convolutional neural networks (CNNs) used for image understanding than unrelated objects (bicycle-fork). Stimuli were of objects against a white background and consisted of a diverse set of contexts (N = 73). CNN representations of contextually related objects were more similar to one another than to unrelated objects across all CNN layers. Critically, the similarity found in CNNs correlated with human behavior across multiple experiments assessing contextual relatedness, emerging significant only in the later layers. The results demonstrate that context is inherently represented in CNNs as a result of object recognition training, and that the representation in the later layers of the network tap into the contextual regularities that predict human behavior.

No abstract

No abstract