We investigated whether the visual hMT+ cortex plays a role in supramodal representation of sensory flow, not mediated by visual mental imagery. We used functional magnetic resonance imaging to measure neural activity in sighted and congenitally blind individuals during passive perception of optic and tactile flows. Visual motion-responsive cortex, including hMT+, was identified in the lateral occipital and inferior temporal cortices of the sighted subjects by response to optic flow. Tactile flow perception in sighted subjects activated the more anterior part of these cortical regions but deactivated the more posterior part. By contrast, perception of tactile flow in blind subjects activated the full extent, including the more posterior part. These results demonstrate that activation of hMT+ and surrounding cortex by tactile flow is not mediated by visual mental imagery and that the functional organization of hMT+ can develop to subserve tactile flow perception in the absence of any visual experience. Moreover, visual experience leads to a segregation of the motion-responsive occipitotemporal cortex into an anterior subregion involved in the representation of both optic and tactile flows and a posterior subregion that processes optic flow only.

Observing and learning actions and behaviors from others, a mechanism crucial for survival and social interaction, engages the mirror neuron system. To determine whether vision is a necessary prerequisite for the human mirror system to develop and function, we used functional magnetic resonance imaging to compare brain activity in congenitally blind individuals during the auditory presentation of hand-executed actions or environmental sounds, and the motor pantomime of manipulation tasks, with that in sighted volunteers, who additionally performed a visual action recognition task. Congenitally blind individuals activated a premotor-temporoparietal cortical network in response to aurally presented actions that overlapped both with mirror system areas found in sighted subjects in response to visually and aurally presented stimuli, and with the brain response elicited by motor pantomime of the same actions. Furthermore, the mirror system cortex showed a significantly greater response to motor familiar than to unfamiliar action sounds in both sighted and blind individuals. Thus, the mirror system in humans can develop in the absence of sight. The results in blind individuals demonstrate that the sound of an action engages the mirror system for action schemas that have not been learned through the visual modality and that this activity is not mediated by visual imagery. These findings indicate that the mirror system is based on supramodal sensory representations of actions and, furthermore, that these abstract representations allow individuals with no visual experience to interact effectively with others.

No abstract

Federated Learning (FL) has emerged as a promising technique for edge devices to collaboratively learn a shared prediction model, while keeping their training data on the device, thereby decoupling the ability to do machine learning from the need to store the data in the cloud. However, FL is difficult to implement and deploy in practice, considering the heterogeneity in mobile devices, e.g., different programming languages, frameworks, and hardware accelerators. Although there are a few frameworks available to simulate FL algorithms (e.g., TensorFlow Federated), they do not support implementing FL workloads on mobile devices. Furthermore, these frameworks are designed to simulate FL in a server environment and hence do not allow experimentation in distributed mobile settings for a large number of clients. In this paper, we present Flower 1 , a FL framework which is both agnostic towards heterogeneous client environments and also scales to a large number of clients, including mobile and embedded devices. Flower's abstractions let developers port existing mobile workloads with little overhead, regardless of the programming language or ML framework used, while also allowing researchers flexibility to experiment with novel approaches to advance the state-of-the-art. We describe the design goals and implementation considerations of Flower and show our experiences in evaluating the performance of FL across clients with heterogeneous computational and communication capabilities. 1 https://flower.dev/ Preprint. Under review.

The present study was designed to investigate the brain functional architecture that subserves visuo-spatial and motor processing in highly skilled individuals. By using functional magnetic resonance imaging (fMRI), we measured brain activity while eleven Formula racing-car drivers and eleven ‘naïve’ volunteers performed a motor reaction and a visuo-spatial task. Tasks were set at a relatively low level of difficulty such to ensure a similar performance in the two groups and thus avoid any potential confounding effects on brain activity due to discrepancies in task execution. The brain functional organization was analyzed in terms of regional brain response, inter-regional interactions and blood oxygen level dependent (BOLD) signal variability. While performance levels were equal in the two groups, as compared to naïve drivers, professional drivers showed a smaller volume recruitment of task-related regions, stronger connections among task-related areas, and an increased information integration as reflected by a higher signal temporal variability. In conclusion, our results demonstrate that, as compared to naïve subjects, the brain functional architecture sustaining visuo-motor processing in professional racing-car drivers, trained to perform at the highest levels under extremely demanding conditions, undergoes both ‘quantitative’ and ‘qualitative’ modifications that are evident even when the brain is engaged in relatively simple, non-demanding tasks. These results provide novel evidence in favor of an increased ‘neural efficiency’ in the brain of highly skilled individuals.