Functional heterogeneity of medial posterior parietal cortex of macaque monkey

The posterior parietal cortex (PPC) of primates represents a remarkable platform that has evolved over time to solve some of the computational challenges that we face in the everyday life, such as sensorimotor integration, spatial attention, and motor planning. With the aim of further investigating the multifaceted functional characteristics of medial PPC, we conducted three studies to explore the visuomotor, somatic, visual, and attention-related properties of two PPC areas: V6A, a visuomotor area part of the dorsomedial visual stream, and PE, an area strongly dominated by somatomotor input, residing mainly on the exposed surface of the superior parietal lobule. In the first study, we tested the impact of visual feedback on V6A grasp-related activity during arm movements towards objects of different shapes. Our results demonstrate that V6A is modulated by both grip type and visual information during grasping preparation and execution, with a predominance of cells influenced by grip type. In the second study, we explored the influence of depth and direction information on reach-related activity of neurons in the so far largely neglected medial part of area PE. We observed a remarkable trend in medial PPC, going from the joint coding of depth and direction signals caudally, in area V6A, to a largely segregated processing of the two signals rostrally, in area PE. In the third study, we used a combined fMRI-electrophysiology experiment to investigate the neuronal mechanisms underlying covert shift of attention processes in area V6A. Our preliminary results reveal that half of the cells showed shift-selective activity when the monkey covertly shifted its attention towards the receptive field. All together these findings highlight the role of the medial PPC in integrating information coming from different sources (vision, somatosensory and motor) and emphasize the involvement of action-related regions of the dorsomedial visual stream in higher level cognitive functions.

Optimizing AI at the Edge: from network topology design to MCU deployment

The first topic analyzed in the thesis will be Neural Architecture Search (NAS). I will focus on two different tools that I developed, one to optimize the architecture of Temporal Convolutional Networks (TCNs), a convolutional model for time-series processing that has recently emerged, and one to optimize the data precision of tensors inside CNNs. The first NAS proposed explicitly targets the optimization of the most peculiar architectural parameters of TCNs, namely dilation, receptive field, and the number of features in each layer. Note that this is the first NAS that explicitly targets these networks. The second NAS proposed instead focuses on finding the most efficient data format for a target CNN, with the granularity of the layer filter. Note that applying these two NASes in sequence allows an "application designer" to minimize the structure of the neural network employed, minimizing the number of operations or the memory usage of the network. After that, the second topic described is the optimization of neural network deployment on edge devices. Importantly, exploiting edge platforms' scarce resources is critical for NN efficient execution on MCUs. To do so, I will introduce DORY (Deployment Oriented to memoRY) -- an automatic tool to deploy CNNs on low-cost MCUs. DORY, in different steps, can manage different levels of memory inside the MCU automatically, offload the computation workload (i.e., the different layers of a neural network) to dedicated hardware accelerators, and automatically generates ANSI C code that orchestrates off- and on-chip transfers with the computation phases. On top of this, I will introduce two optimized computation libraries that DORY can exploit to deploy TCNs and Transformers on edge efficiently. I conclude the thesis with two different applications on bio-signal analysis, i.e., heart rate tracking and sEMG-based gesture recognition.