Vertical Federated Learning (VFL) enables multiple data owners to collaboratively train a global model, while preserving data privacy. However, significant communication overhead arises during the training phase due to the transmission of massive encrypted intermediate results. Our investigation into the performance of the model with varying participant numbers reveals that increasing participants, especially in larger groups, may not necessarily yield substantial performance improvements. Instead, it tends to introduce additional communication overhead. Therefore, in the context of VFL, participant selection emerges as a crucial strategy to mitigate communication cost, though it remains a relatively unexplored domain. In this work, we propose a novel adaptive participant selection method in VFL, named VFLMG, to dynamically determine the number of selected participants, as well as to identify the group of participants that best strikes a balance between model performance and communication overhead. Evaluations on various datasets show that VFLMG significantly reduces communication overhead while only causing a minimal loss in accuracy. For instance, in the MNIST and Fashion MNIST datasets, VFLMG reduces communication costs by 32.1 % and 27.6 % respectively, with an accuracy drop of 0.5 % and 0.3 %.

This paper conducts a systematic study on the role of visual attention in Unsupervised Video Object Segmentation (UVOS) tasks. By elaborately annotating three popular video segmentation datasets (DAVIS, Youtube-Objects and SegTrack V2) with dynamic eye-tracking data in the UVOS setting, for the first time, we quantitatively verified the high consistency of visual attention behavior among human observers, and found strong correlation between human attention and explicit primary object judgements during dynamic, task-driven viewing. Such novel observations provide an in-depth insight into the underlying rationale behind UVOS. Inspired by these findings, we decouple UVOS into two sub-tasks: UVOS-driven Dynamic Visual Attention Prediction (DVAP) in spatiotemporal domain, and Attention-Guided Object Segmentation (AGOS) in spatial domain. Our UVOS solution enjoys three major merits: 1) modular training without using expensive video segmentation annotations, instead, using more affordable dynamic fixation data to train the initial video attention module and using existing fixation-segmentation paired static/image data to train the subsequent segmentation module; 2) comprehensive foreground understanding through multi-source learning; and 3) additional interpretability from the biologically-inspired and assessable attention. Experiments on popular benchmarks show that, even without using expensive video object mask annotations, our model achieves compelling performance in comparison with state-of-the-arts.

Clinical data warehouse has been developed as a fundamental data infrastructure for large scale TCM clinical data management and decision support services. However, as a key component, data extraction, transforming and loading (ETL) is a complicated and labor intensive task to ensure high data quality before all kinds of data analyses. This paper introduces an enhanced ETL technique framework, which includes operational data store (ODS) model and two step data preprocessing subcomponents, to perform the ETL tasks. The ODS data model was designed to integrate the heterogeneous clinical data sources and support the direct copy from these data sources to ODS database by ETL. Therefore, ETL task has been separated into two core steps in enhanced ETL component: (1) dynamic filter and copy of the original operational data sources to ODS; (2) specialized transforming the ODS data to detailed clinical data warehouse. This enhanced technique framework improves the ETL performance to be used in clinical data center since there would have various kinds of operational data sources that need be integrated in this data environments. This paper has a description of the related enhanced ETL framework and proposes some key procedures to accomplish the tasks.

A lossless compression algorithm for synthetic aperture radar (SAR) amplitude images based on modified quadtree bit-plane coding is proposed. First, a SAR amplitude image is divided into independent bit planes, and then, the probability of bit “0” is computed and compared with the predefined threshold in order to select the optimal block size for each bit plane. Finally, a modified quadtree coding method is adopted to encode each block data. Additionally, the computing method of predefined threshold is also addressed in this letter. The experimental results show that the proposed method outperforms previous methods for all SAR images of the test sets; an average 0.895-bpp decrease in bit rate was observed when compared with JPEG-LS.