Pablo Muñoz

Julia Caleya-Sanchez, Pablo Muñoz, Jorge Sánchez-Garrido et al.

Time synchronization is essential for industrial IoT and Industry 4.0/5.0 applications, but achieving high synchronization accuracy in Time-Sensitive Networking (TSN)-5G networks is challenging due to jitter and asymmetric delays. 3GPP TS 23.501 defines three 5G synchronization modes: time-aware system, boundary clock (BC), and transparent clock (TC), where TC offers a promising solution. However, to the best of our knowledge, there is no empirical evaluation of TC in a TSN-5G network. This paper empirically evaluates an 5G end-to-end TC in a TSN-5G network, implemented on commercial TSN switches with a single clock. For TC development, we compute the residence time in 5G and recover the clock domain at the slave node. We deploy a TSN-5G testbed with commercial equipment for synchronization evaluation by modifying the Precision Timing Protocol (PTP) message transmission rates. Experimental results show a peak-to-peak synchronization of 500 ns, meeting the industrial requirement of < 1 us, with minimal synchronization offsets for specific PTP message transmission rates.

Waqwoya Abebe, Sadegh Jafari, Sixing Yu et al.

Neural Architecture Search (NAS) is a powerful approach of automating the design of efficient neural architectures. In contrast to traditional NAS methods, recently proposed one-shot NAS methods prove to be more efficient in performing NAS. One-shot NAS works by generating a singular weight-sharing supernetwork that acts as a search space (container) of subnetworks. Despite its achievements, designing the one-shot search space remains a major challenge. In this work we propose a search space design strategy for Vision Transformer (ViT)-based architectures. In particular, we convert the Segment Anything Model (SAM) into a weight-sharing supernetwork called SuperSAM. Our approach involves automating the search space design via layer-wise structured pruning and parameter prioritization. While the structured pruning applies probabilistic removal of certain transformer layers, parameter prioritization performs weight reordering and slicing of MLP-blocks in the remaining layers. We train supernetworks on several datasets using the sandwich rule. For deployment, we enhance subnetwork discovery by utilizing a program autotuner to identify efficient subnetworks within the search space. The resulting subnetworks are 30-70% smaller in size compared to the original pre-trained SAM ViT-B, yet outperform the pretrained model. Our work introduces a new and effective method for ViT NAS search-space design.