Low Earth orbit (LEO) satellites nowadays play a pivotal role in collecting images for the Earth observation. However, the images collected by satellites are possibly tremendous, which causes challenges in dealing with the satellite images. Those challenges include: 1) the unrealistic of transmitting those massive image data to the ground station for centralized analysis because of restricted satellite communication bandwidth and the data privacy issue, and 2) satellite data may be non-independent and identically distributed (non-IID). In this paper, we propose a prototype-guided federated knowledge distillation (Pro-FedKD) approach in an LEO Satellite-high altitude platform (HAP) system, which is designed based on self-knowledge distillation (SKD), federated prototype learning (FedProto) and federated learning (FL). Owing to the adoption of FL, the first challenge can be handled since FL does not require data to leave the local side. To cope with the second challenge, SKD and FedProto are employed. In addition, both model aggregation and prototype aggregation are employed on a pre-defined HAP. To enhance the effectiveness, a top-$N$model aggregation mechanism is proposed, in which among all models,$N$local models that can achieve the top$N$maximum accuracies over the validation dataset of the pre-defined HAP will be selected for aggregation. Experiments demonstrate the error rate gained by the proposed Pro-FedKD method is separately 3.76×, 3.17×, 1.55×, and 1.18× smaller than FedExP, MOON, FedProto, and pFedSD over the EuroSAT dataset, demonstrating a significant reduction. The proposed method also exhibits preeminence in other datasets.

RGB-based 3D hand pose estimation has been successful for decades thanks to large-scale databases and deep learning. However, the hand pose estimation network does not operate well for hand pose images whose characteristics are far different from the training data. This is caused by various factors such as illuminations, camera angles, diverse backgrounds in the input images, etc. Many existing methods tried to solve it by supplying additional large-scale unconstrained/target domain images to augment data space; however collecting such large-scale images takes a lot of labors. In this paper, we present a simple image-free domain generalization approach for the hand pose estimation framework that uses only source domain data. We try to manipulate the image features of the hand pose estimation network by adding the features from text descriptions using the CLIP (Contrastive Language-Image Pretraining) model. The manipulated image features are then exploited to train the hand pose estimation network via the contrastive learning framework. In experiments with STB and RHD datasets, our algorithm shows improved performance over the state-of-the-art domain generalization approaches.

We propose WALTZ, an LSM tree-based key-value store on the emerging Zoned Namespace (ZNS) SSD. The key contribution of WALTZ is to leverage the zone append command, which is a recent addition to ZNS SSD specifications, to provide tight tail latency. The long tail latency problem caused by the merging process of multiple parallel writes, called batch-group writes, is effectively addressed by the internal synchronization mechanism of ZNS SSD. To provide fast failover when the active zone becomes full for a write-ahead log (WAL) file during parallel append, WALTZ introduces a mechanism for WAL zone replacement and reservation. Finally, lazy metadata management allows a put query to be processed fast without requiring any other synchronizations to enable lock-free execution of individual append commands. For evaluation we use both mi-crobenchmarks (db_bench) with varying read/write ratios and key skewnesses, and realistic social-graph workloads (MixGraph from Facebook). Our evaluation demonstrates geomean reduction of tail latency by 2.19× and 2.45× for db_bench and MixGraph, respectively, with a maximum reduction of 3.02× and 4.73×. As a side effect of eliminating the overhead of batch-group writes, WALTZ also improves the query throughput (QPS) by up to 11.7%.

In this era of sophisticated technology for smart cities, when communication between smart things is crucial, Internet of Everything (IoE) networks play a key role in merging the cyber and physical worlds. IoE networks are used in a range of applications, including smart agriculture, smart housing, and smart medical services, thanks to the implementation of smart sensor networks (SSN). However, if human administration of the IoE network is impossible due to unforeseen reasons, the installed IoE network's life cycle is critical. The ability to reduce the amount of energy consumed by an IoE network is crucial for extending the network's life cycle. The objective of this work is to tackle the tough task of reducing IoE network energy usage (EU) on a wide scale. This study proposed an energy-efficient IoE network deployment problem for SSN, unmanned aerial vehicles (UAVs), and low earth orbit (LEO) satellites to achieve the aim of energy-efficient utilization of the stored UAVs' energy. For managing the EU of the IoE network, the proposed problem is a mixed-integer linear programming (MILP) optimization problem which is NP-hard in nature. To address this challenge, we use genetic algorithms (GA) to solve the task of minimizing EU while maintaining a low level of complexity. The proposed solution to the EU problem is flexible and effective, and it contributes to the IoE network's goal of a low EU and a long system lifespan.

To meet surging demands for deep learning inference services, many cloud computing vendors employ high-performance specialized accelerators, called neural processing units (NPUs). One important challenge for effective use of NPUs is to achieve high resource utilization over a wide spectrum of deep neural network (DNN) models with diverse arithmetic intensities. There is often an intrinsic mismatch between the compute-to-memory bandwidth ratio of an NPU and the arithmetic intensity of the model it executes, leading to under-utilization of either compute resources or memory bandwidth. Ideally, we want to saturate both compute TOP/s and DRAM bandwidth to achieve high system throughput. Thus, we propose Layerweaver, an inference serving system with a novel multi-model time-multiplexing scheduler for NPUs. Layerweaver reduces the temporal waste of computation resources by interweaving layer execution of multiple different models with opposing characteristics: compute-intensive and memory-intensive. Layerweaver hides the memory time of a memory-intensive model by overlapping it with the relatively long computation time of a compute-intensive model, thereby minimizing the idle time of the computation units waiting for off-chip data transfers. For a two-model serving scenario of batch 1 with 16 different pairs of compute- and memory-intensive models, Layerweaver improves the temporal utilization of computation units and memory channels by 44.0% and 28.7%, respectively, to increase the system throughput by 60.1% on average, over the baseline executing one model at a time.