Split Federated Learning (SFL) splits and collaboratively trains a shared model between clients and server, where clients transmit activations and client-side models to server for updates. Recent SFL studies assume synchronous transmission of activations and client-side models from clients to server. However, due to significant variations in computational and communication capabilities among clients, activations and client-side models arrive at server asynchronously. The delay caused by asynchrony significantly degrades the performance of SFL. To address this issue, we consider an asynchronous SFL framework, where an activation buffer and a model buffer are embedded on the server to manage the asynchronously transmitted activations and client-side models, respectively. Furthermore, as asynchronous activation transmissions cause the buffer to frequently receive activations from resource-rich clients, leading to biased updates of the server-side model, we propose Generative activations-aided Asynchronous SFL (GAS). In GAS, the server maintains an activation distribution for each label based on received activations and generates activations from these distributions according to the degree of bias. These generative activations are then used to assist in updating the server-side model, ensuring more accurate updates. We derive a tighter convergence bound, and our experiments demonstrate the effectiveness of the proposed method.

Point cloud registration is a fundamental yet challenging task in computer vision and robotics. While framing it as a reconstruction problem has shown promise, traditional reconstruction approaches rely on positional encodings to encode positional information, which inadequately capture the intricate geometric and positional relationships between point cloud pairs, leading to suboptimal registration results. To overcome this limitation, we introduce a Local Geometry Information Aggregation (LGIA) module that effectively captures both global positional context and fine-grained local geometric details. Furthermore, to better utilize the geometric and positional information inherent in point clouds, we propose a salient point sampling strategy that increases the proportion during the downsampling process. Subsequently, we assign these salient points higher weights during patch matching. Experiments demonstrate the superiority of our method on 3DMatch, 3DLoMatch and KITTI datasets, which achieves SOTA results at three metrics of FRM, RRT and RTE, and especially 98.5% FRM on the 3DMatch dataset.

Split Federated Learning (SFL) is a distributed machine learning framework where the models are split and trained on the server and clients. However, data heterogeneity and partial client participation result in label distribution skew, which severely degrades learning performance. To address this issue, we propose Concatenated Activations Enabled SFL with Logit Adjustments, in which activations from client-side models are concatenated as the input of the server-side model to centrally adjust label distribution across different clients, and logit adjustments in the loss functions of both server-side and client-side models are performed to deal with the label distribution variation across different subsets of participating clients. Experiments demonstrate the superiority of the proposed method compared with the traditional schemes.

This paper focuses on the multi-splitting of DNN over wireless multi-hop networks to distribute the computing over multiple network nodes for achieving efficient edge infer-ence. The challenge is how to choose an inference routing in which both the transmission and inference can be efficiently relayed hop-by-hop. We propose a DNN multi-splitting method along with dynamical early-exit of inference based on communication and computation conditions. We formulate and then solve an optimization problem of joint routing, split points selection, and model deployment to minimize the end - to-end inference latency. The experimental results demonstrate the superiority of our work in reducing inference latency.

Deep learning-based methods show promise for detecting crop pests and diseases, but significant challenges remain in moth detection. These challenges stem from the significant morphological variations of moths across different life stages and the agility exhibited by adult moths. In this paper, an improved MAG-YOLOv8 model based on YOLOv8 is proposed for the detection of moth adults and larvae. The main innovations of this model are summarized as follows: (1) Multi-scale convolution structure (MSCS) is introduced into the backbone network to enhance the feature extraction capability; (2) The Attentional Scale Sequence Fusion (ASF-P2) module is incorporated into the feature fusion network to enhance the detection accuracy for small targets; (3) The Programmable Gradient Information (PGI) module is introduced to enhance the training process efficiency while preserving reasoning efficacy via reparameterization techniques. Furthermore, based on 1,600 manually labeled images, data augmentation was used to build a dataset of eight moth species in both adult and larval stages, with 500 images per stage, totaling 8,000 samples for training. Experimental results show that MAG-YOLOv8 achieves high accuracy, with Precision, Recall, mAP50, and mAP50-95 reaching 93.2%, 88.5%, 94.3%, and 69.5% respectively. These metrics are 2.3%, 4.5%, 2.9%, and 4.5% higher than those of the benchmark YOLOv8 model. The proposed MAG-YOLOv8 model demonstrated superior performance in detecting moth adults and larvae across diverse agricultural environments, and provided a practical solution for automated pest monitoring and management.

Federated learning (FL) is a promising distributed learning framework where distributed clients collaboratively train a machine learning model coordinated by a server. To tackle the stragglers issue in asynchronous FL, we consider that each client keeps local updates and probabilistically transmits the local model to the server at arbitrary times. We first derive the (approximate) expression for the convergence rate based on the probabilistic client selection. Then, an optimization problem is formulated to trade off the convergence rate of asynchronous FL and mobile energy consumption by joint probabilistic client selection and bandwidth allocation. We develop an iterative algorithm to solve the non-convex problem globally optimally. Experiments demonstrate the superiority of the proposed approach compared with the traditional schemes.

Distributed Stein Variational Gradient Descent (DSVGD) is a non-parametric distributed learning framework for federated Bayesian learning, where multiple clients jointly train a machine learning model by communicating a number of non-random and interacting particles with the server. Since communication resources are limited, selecting the clients with most informative local learning updates can improve the model convergence and communication efficiency. In this paper, we propose two selection schemes for DSVGD based on Kernelized Stein Discrepancy (KSD) and Hilbert Inner Product (HIP). We derive the upper bound on the decrease of the global free energy per iteration for both schemes, which is then minimized to speed up the model convergence. We evaluate and compare our schemes with conventional schemes in terms of model accuracy, convergence speed, and stability using various learning tasks and datasets.