Multiple Datasets Research Articles

STG-MTL: Scalable Task Grouping For Multi-Task Learning Using Data Maps

Abstract Multi-Task Learning (MTL) is a powerful technique that has gained popularity due to its performance improvement over traditional Single-Task Learning (STL). However, MTL is often challenging because there is an exponential number of possible task groupings, which can make it difficult to choose the best one because some groupings might produce performance degradation due to negative interference between tasks. That is why existing solutions are severely suffering from scalability issues, limiting any practical application. In our paper, we propose a new data-driven method that addresses these challenges and provides a scalable and modular solution for classification task grouping based on a re-proposed data-driven features, Data Maps, which capture the training dynamics for each classification task during the MTL training. Through a theoretical comparison with other techniques, we manage to show that our approach has the superior scalability. Our experiments show a better performance and verify the method's effectiveness, even on an unprecedented number of tasks (up to 100 tasks on CIFAR100). Being the first to work on such number of tasks, our comparisons on the resulting grouping shows similar grouping to the mentioned in the dataset, CIFAR100. Finally, we provide a modular implementation for easier integration and testing, with examples from multiple datasets and tasks.

FedUB: Federated Learning Algorithm Based on Update Bias

Federated learning, as a distributed machine learning framework, aims to protect data privacy while addressing the issue of data silos by collaboratively training models across multiple clients. However, a significant challenge to federated learning arises from the non-independent and identically distributed (non-iid) nature of data across different clients. non-iid data can lead to inconsistencies between the minimal loss experienced by individual clients and the global loss observed after the central server aggregates the local models, affecting the model’s convergence speed and generalization capability. To address this challenge, we propose a novel federated learning algorithm based on update bias (FedUB). Unlike traditional federated learning approaches such as FedAvg and FedProx, which independently update model parameters on each client before direct aggregation to form a global model, the FedUB algorithm incorporates an update bias in the loss function of local models—specifically, the difference between each round’s local model updates and the global model updates. This design aims to reduce discrepancies between local and global updates, thus aligning the parameters of locally updated models more closely with those of the globally aggregated model, thereby mitigating the fundamental conflict between local and global optima. Additionally, during the aggregation phase at the server side, we introduce a metric called the bias metric, which assesses the similarity between each client’s local model and the global model. This metric adaptively sets the weight of each client during aggregation after each training round to achieve a better global model. Extensive experiments conducted on multiple datasets have confirmed the effectiveness of the FedUB algorithm. The results indicate that FedUB generally outperforms methods such as FedDC, FedDyn, and Scaffold, especially in scenarios involving partial client participation and non-iid data distributions. It demonstrates superior performance and faster convergence in tasks such as image classification.

Underwater image enhancement via color correction and multi-feature image fusion

Abstract The light attenuation underwater causes the actual underwater images to suffer from color cast, low contrast, and weak illumination. To address these issues, an effective fusion-based method is proposed, which realizes color correction, brightness adjustment, contrast, and detail enhancement of underwater images. Concretely, we first design an adaptive color correction method via dominant color channel judgment and lower color channel compensation. Then, we detect the brightness of each input image and propose a gamma correction function based on the gradient of the cumulative histogram to adjust the brightness of the low-light images. Subsequently, global histogram stretching and adaptive fractional differentiation techniques are employed to process the brightness-adjusted image, and then the global contrast-enhanced version and detail-enhanced version are generated respectively. To integrate the advantages of both versions, a channel fusion method based on the Lab color space is used to fuse the luminance and color of the two versions separately. The experimental results demonstrate the effectiveness of the proposed method in improving the color and illumination of underwater images, as well as enhancing the clarity of images. Moreover, the testing results on multiple datasets validate the excellent stability of this method.

A Robust Coverless Image Steganography Algorithm Based on Image Retrieval with SURF Features

With the advancement of image steganography, coverless image steganography has gained widespread attention due to its ability to hide information without modifying the carrier of images. However, existing coverless image steganography methods often require both communicating parties to transmit an amount of additional information including image blocks’ locations or a large number of parameters, which will raise a serious suspicion. In light of this issue, we propose a robust coverless image steganography algorithm based on Speeded-Up Robust Features (SURF). Firstly, the proposed method allows both communicating parties to independently create multiple coverless image datasets (CIDs) using random seeds. Then, a mapping rule is designed for creating one-to-one correspondence between hash sequences and images in CIDs. Finally, the secret information will be carried by the images whose hash sequences are equal to the secret segments. At the receiver side, the robust SURF of images is utilized to retrieve the secret information. Experimental results demonstrate that the proposed algorithm outperforms other methods in terms of capacity, robustness, and security. Furthermore, it is worth noting that the proposed method eliminates the need to transmit a large amount of additional information, which is a significant security issue in existing coverless image steganography algorithms.

Node classification in complex networks based on multi-view debiased contrastive learning

AbstractIn complex networks, contrastive learning has emerged as a crucial technique for acquiring discriminative representations from graph data. Maximizing the similarity among relevant sample pairs while minimizing that among irrelevant pairs is pivotal in contrastive learning. Therefore, careful consideration must be given to the design of sample pairs in contrastive learning. However, existing node-level self-supervised contrastive learning often treats the enhanced representation of a central node as a positive sample, while considering representations of all other nodes as negative samples. This approach can lead to conflicts in downstream tasks on some graph data, as nodes of the same class are treated as negatives during learning. Precision in sample pair design is essential for enhancing the performance of contrastive learning. To address this issue, this paper introduces a negative sample debiased sampling contrastive learning (NDSCL), specifically tailored for node classification tasks. In particular, this method integrates contrastive learning with semi-supervised learning. A trained classifier assigns pseudo-labels to unlabeled data, and debiased sampling is applied to negative samples. Unlike other methods that focus on negative sample selection, NDSCL also addresses the imbalance in pseudo-label distribution by employing debiasing techniques. Finally, in conjunction with diffusion augmentation, the model is provided with diverse views as inputs to maximize the retention of underlying semantic information. Experimental results demonstrate that the proposed model significantly outperforms baseline models in node-level classification tasks across multiple network datasets. Moreover, the model not only enhances accuracy but also improves computational speed and memory requirements for handling large-scale graph data structures.

A longer time to relapse is associated with a larger increase in differences between paired primary and recurrent IDH wild-type glioblastomas at both the transcriptomic and genomic levels

Glioblastoma (GBM) is the most common malignant brain tumor in adults, which remains incurable and often recurs rapidly after initial therapy. While large efforts have been dedicated to uncover genomic/transcriptomic alternations associated with the recurrence of GBMs, the evolutionary trajectories of matched pairs of primary and recurrent (P-R) GBMs remain largely elusive. It remains challenging to identify genes associated with time to relapse (TTR) and construct a stable and effective prognostic model for predicting TTR of primary GBM patients. By integrating RNA-sequencing and genomic data from multiple datasets of patient-matched longitudinal GBMs of isocitrate dehydrogenase wild-type (IDH-wt), here we examined the associations of TTR with heterogeneities between paired P-R GBMs in gene expression profiles, tumor mutation burden (TMB), and microenvironment. Our results revealed a positive correlation between TTR and transcriptomic/genomic differences between paired P-R GBMs, higher percentages of non-mesenchymal-to-mesenchymal transition and mesenchymal subtype for patients with a short TTR than for those with a long TTR, a high correlation between paired P-R GBMs in gene expression profiles and TMB, and a negative correlation between the fitting level of such a paired P-R GBM correlation and TTR. According to these observations, we identified 55 TTR-associated genes and thereby constructed a seven-gene (ZSCAN10, SIGLEC14, GHRHR, TBX15, TAS2R1, CDKL1, and CD101) prognostic model for predicting TTR of primary IDH-wt GBM patients using univariate/multivariate Cox regression analyses. The risk scores estimated by the model were significantly negatively correlated with TTR in the training set and two independent testing sets. The model also segregated IDH-wt GBM patients into two groups with significantly divergent progression-free survival outcomes and showed promising performance for predicting 1-, 2-, and 3-year progression-free survival rates in all training and testing sets. Our findings provide new insights into the molecular understanding of GBM progression at recurrence and potential targets for therapeutic treatments.

Satellite-Based Estimation of Near-Surface NO2 Concentration in Cloudy and Rainy Areas

Satellite-based remote sensing enables the quantification of tropospheric NO2 concentrations, offering insights into their environmental and health impacts. However, remote sensing measurements are often impeded by extensive cloud cover and precipitation. The scarcity of valid NO2 observations in such meteorological conditions increases data gaps and thus hinders accurate characterization and variability of concentration across geographical regions. This study utilizes the Empirical Orthogonal Function interpolation in conjunction with the Extreme Gradient Boosting (XGBoost) algorithm and dense urban atmospheric observed station data to reconstruct continuous daily tropospheric NO2 column concentration data in cloudy and rainy areas and thereby improve the accuracy of NO2 concentration mapping in meteorologically obscured regions. Using Chengdu City as a case study, multiple datasets from satellite observations (TROPOspheric Monitoring Instrument, TROPOMI), near-surface NO2 measurements, meteorology, and ancillary data are leveraged to train models. The results showed that the integration of reconstructed satellite observations with provincial and municipal control surface measurements enables the XGBoost model to achieve heightened predictive accuracy (R2 = 0.87) and precision (RMSE = 5.36 μg/m3). Spatially, this approach effectively mitigates the problem of missing values in estimation results due to absent satellite data while simultaneously ensuring increased consistency with ground monitoring station data, yielding images with more continuous and refined details. These results underscore the potential for reconstructing satellite remote sensing information and combining it with dense ground observations to greatly improve NO2 mapping in cloudy and rainy areas.

Robust anomaly detection in industrial images by blending global–local features

AbstractIndustrial image anomaly detection achieves automated detection and localization of defects or abnormal regions in images through image processing and deep learning techniques. Currently, utilizing the approach of reverse knowledge distillation has yielded favourable outcomes. However, it is still a challenge in terms of the feature extraction capability of the image and the robustness of the decoding of the student network. This study first addresses the issue that the teacher network has not been able to extract global information more effectively. To acquire more global information, a vision transformer network is introduced to enhance the model's global information extraction capability, obtaining better features to further assist the student network in decoding. Second, for anomalous samples, to address the low similarity between features extracted by the teacher network and features restored by the student network, Gaussian noise is introduced. This further increases the probability that the features decoded by the student model match normal sample features, enhancing the robustness of the student model. Extensive experiments were conducted on industrial image datasets AeBAD, MvtecAD, and BTAD. In the AeBAD dataset, under the PRO performance metric, the result is 89.83%, achieving state‐of‐the‐art performance. Under the AUROC performance metric, it reaches 83.35%. Similarly, good results were achieved on the MvtecAD and BTAD datasets. The proposed method's effectiveness and performance advantages were validated across multiple industrial datasets, providing a valuable reference for the application of industrial image anomaly detection methods.

Ocean Heat Convergence and North Atlantic multidecadal heat content variability

Abstract We construct an upper ocean (0-1000m) North Atlantic heat budget (26°-67°N) for the period 1950-2020 using multiple observational datasets and an eddy-permitting global ocean model. On multidecadal timescales ocean heat transport convergence controls ocean heat content (OHC) tendency in most regions of the North Atlantic with little role for diffusive processes. In the subpolar North Atlantic (45°N-67°N) heat transport convergence is explained by geostrophic currents whereas ageostrophic currents make a significant contribution in the subtropics (26°N-45°N). The geostrophic contribution in all regions is dominated by anomalous advection across the time-mean temperature gradient although other processes make a significant contribution particularly in the subtropics. The timescale and spatial distribution of the anomalous geostrophic currents are consistent with a simple model of basin scale thermal Rossby waves propagating westwards/northwestwards in the subpolar gyre and multidecadal variations in regional OHC are explained by geostrophic currents periodically coming into alignment with the mean temperature gradient as the Rossby wave passes through. The global ocean model simulation shows that multidecadal variations in the Atlantic Meridional Overturning Circulation are synchronized with the ocean heat transport convergence consistent with modulation of the west-east pressure gradient by the propagating Rossby wave.

ADHD classification with cross-dataset feature selection for biomarker consistency detection

Objective. Attention deficit hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder in children. While numerous intelligent methods are applied for its subjective diagnosis, they seldom consider the consistency problem of ADHD biomarkers. In practice, these data-driven approaches lead to varying learned features for ADHD classification across diverse ADHD datasets. This phenomenon significantly undermines the reliability of identified biomarkers and hampers the interpretability of these methods. Approach. In this study, we propose a cross-dataset feature selection (FS) module using a grouped SVM-based recursive feature elimination approach (G-SVM-RFE) to enhance biomarker consistency across multiple datasets. Additionally, we employ connectome gradient data for ADHD classification. In details, we introduce the G-SVM-RFE method to effectively concentrate gradient components within a few brain regions, thereby increasing the likelihood of identifying these regions as ADHD biomarkers. The cross-dataset FS module is integrated into an existing binary hypothesis testing (BHT) framework. This module utilizes external datasets to identify global regions that yield stable biomarkers. Meanwhile, given a dataset which waits for implementing the classification task as local dataset, we learn its own specific regions to further improve the performance of accuracy on this dataset. Main results. By employing this module, our experiments achieve an average accuracy of 96.7% on diverse datasets. Importantly, the discriminative gradient components primarily originate from the global regions, providing evidence for the significance of these regions. We further identify regions with the high appearance frequencies as biomarkers, where all the used global regions and one local region are recognized. Significance. These biomarkers align with existing research on impaired brain regions in children with ADHD. Thus, our method demonstrates its validity by providing enhanced biological explanations derived from ADHD mechanisms.

English Speaking Assessment Algorithm Based on Deep Learning

English as a Foreign Language (EFL) students perform when speaking in public. An increasingly globalized world, effective public speaking is critical, but EFL students struggle to perform it, despite importance of qualities such as eye contact, speech pauses, there is presently no objective examination of such elements. A summative assessment has historically been the predominant form of evaluation in college English speaking assessments. Exam-centric teaching has considerable negative effect on foreign language training. In this research work, English Speaking Assessment Algorithm Based on Deep Learning (ESA-NEGCN-NBOA) is proposed. Initially, input video data are gathered from the multiple video dataset (MVD).The input video data is then pre-processed using Deep Attentional Guided Image Filtering (DAGIF) to remove presence of signal-dependent noise and improve lack of pixels from the regions and enhanced the video data. The data that has been pre-processed is utilized to Feature extraction using New General Double Integral Transform (NGDIT), which extract the significant attributes such as mel-frequency cepstral coefficients, energy, speech rate and pitch. Then NEGCN is proposed to improve students spoken English performance by assessmenting the English speakers. In general, NEGCN doesn’t express some adaption of optimization approaches for determining optimal parameters to promise exact improvement of assessment. Therefore, NBOA is proposed to enhance weight parameter of NEGCN for English speaking assessments, which precisely assess the English speaking. Performance measures such as accuracy, assessment error, evaluation time, pretest and posttest are examined when the proposed ESA-NEGCN-NBOA method is put into practice. The proposed ESA-NEGCN-NBOA method attains21.36%, 23.42% and 19.29% higher accuracy, 23.36%, 18.42% and 28.27% lower evaluation error, 20.36%, 27.42%, 28.17% lesser evaluation time analysed with existing techniques, likes innovative strategy towards oral English assessment utilizing machine learning, data mining, blockchain methods(IST-OEA-ML), machine learning assessment system for spoken English depend on linear predictive coding (AS-SE-LPC-ML), multimodal transfer learning for oral presentation assessment (MM-TL-OPA) respectively.

Structural framework of the Caswell Sub-basin, North West Shelf, Australia

The Caswell Sub-basin, situated within the Browse Basin in the North West Shelf constitutes one of Australia’s primary hydrocarbon producing regions, with notable gas-condensate producing fields including Ichthys and Prelude. Jurassic syn-rift sandstones are extensively distributed across the basin and serve as one of the major reservoirs. However, reservoir sequences are typically intensely faulted and exhibit heterogeneity in thickness and lithofacies, with some areas experiencing localised erosion on uplifted fault blocks (e.g. Northern Caswell Sub-basin). Hence, understanding the nature of the 3D fault patterns and their growth history is crucial for evaluating the reservoir characteristics for field development and additional exploration activities. This study, therefore, aims to evaluate the structural framework and the tectonic evolution of the Caswell Sub-basin. Detailed structural interpretation of Paleozoic and younger sequences was conducted using multiple 3D seismic datasets extending over Ichthys, Prelude, Lasseter, and Crown fields. Our study focuses mainly on Mesozoic faulting patterns and kinematics evaluated from interpreted structural maps, thickness changes of each stratigraphic interval and fault throw profiles of major bounding faults. The extensional phase during the Early–Middle Jurassic triggered the development of NE–SW trending faults and the deepening of the Brewster Graben. While, from the Late Jurassic to the Early Cretaceous, the development of E–W trending faults in the north of the Crown and Lasseter fields indicate a shift in the regional stress regime. We highlight the importance of evaluating the structural linkage from basement to cover sequences to achieve a comprehensive understanding of reservoirs and associated petroleum systems.

Integrated Bioinformatics Analysis Reveals ceRNA Triplet Related to Apoptosis in Gastric Cancer

Apoptosis is a form of programmed cell death and evading the apoptosis is a milestone in gastric cancer (GC) tumorigenesis. Various RNAs such as miRNA (microRNA), mRNA (messenger RNA), lncRNA (long non-coding RNA), circRNA (circular RNA), play crucial roles in the apoptosis mechanism. In this present study, we aimed to understand the role played by these RNAs in apoptosis in GC. We constructed a ceRNA (competitive endogenous RNA) network using the expression profiles obtained from SRA (Sequence Read Archive) and GEO (Gene Expression Omnibus) datasets. After network construction, the differentially expressed mRNAs were used for gene prioritization. The prioritized genes and the RNAs interacting with them were analyzed to study their role in GC apoptosis. Then, we performed functional and pathway analysis to understand the role played by these genes in gastric cancer. This resulted in 37 miRNA-mRNA interactions, one miRNA-lncRNA interaction, and 17 miRNA-circRNA interactions. The binding site analysis resulted in 10 miRNAs that share common MRE (miRNA Response Elements) among mRNA, lncRNAs, circRNA. Besides, we found 33 miRNA-mRNA-circRNA and two miRNA-mRNA-lncRNA valid interactions. Integration of multiple omics datasets revealed dysregulated genes, including IGF1, MME, CCND2, CDH2, and COL1A1, implicated in GC apoptosis. Functional enrichment analysis highlighted pathways related to apoptosis, with CCND2 emerging as a key player. The integrative methodology has revealed a new potential diagnostic biomarker for the regulation of gastric cancer (GC) apoptosis: the CCND2-miR-141-3p-hsa_circ_0008035 ceRNA triplet. This discovery offers fresh perspectives into the intricate regulatory pathways governing gene expression in GC, promising significant insights into its pathology.

FetchEEG: a hybrid approach combining feature extraction and temporal-channel joint attention for EEG-based emotion classification

Objective. Electroencephalogram (EEG) analysis has always been an important tool in neural engineering, and the recognition and classification of human emotions are one of the important tasks in neural engineering. EEG data, obtained from electrodes placed on the scalp, represent a valuable resource of information for brain activity analysis and emotion recognition. Feature extraction methods have shown promising results, but recent trends have shifted toward end-to-end methods based on deep learning. However, these approaches often overlook channel representations, and their complex structures pose certain challenges to model fitting. Approach. To address these challenges, this paper proposes a hybrid approach named FetchEEG that combines feature extraction and temporal-channel joint attention. Leveraging the advantages of both traditional feature extraction and deep learning, the FetchEEG adopts a multi-head self-attention mechanism to extract representations between different time moments and channels simultaneously. The joint representations are then concatenated and classified using fully-connected layers for emotion recognition. The performance of the FetchEEG is verified by comparison experiments on a self-developed dataset and two public datasets. Main results. In both subject-dependent and subject-independent experiments, the FetchEEG demonstrates better performance and stronger generalization ability than the state-of-the-art methods on all datasets. Moreover, the performance of the FetchEEG is analyzed for different sliding window sizes and overlap rates in the feature extraction module. The sensitivity of emotion recognition is investigated for three- and five-frequency-band scenarios. Significance. FetchEEG is a novel hybrid method based on EEG for emotion classification, which combines EEG feature extraction with Transformer neural networks. It has achieved state-of-the-art performance on both self-developed datasets and multiple public datasets, with significantly higher training efficiency compared to end-to-end methods, demonstrating its effectiveness and feasibility.

Retrosynthesis Zero: Self-Improving Global Synthesis Planning Using Reinforcement Learning.

The field of computer-aided synthesis planning (CASP) has witnessed significant growth in recent years. Still, many CASP programs rely on large data sets to train neural networks, resulting in limitations due to the data quality and prior knowledge from chemists. In response, we propose Retrosynthesis Zero (ReSynZ), a reaction template-based method that combines Monte Carlo Tree Search with reinforcement learning inspired by AlphaGo Zero. Unlike other single-step reaction template-based CASP methods, ReSynZ takes complete synthesis paths for complex molecules, determined by reaction rules, as input for training the neural network. ReSynZ enables neural networks trained with relatively small reaction data sets (tens of thousands of data) to generate multiple synthesis pathways for a target molecule and suggest possible reaction conditions. On multiple data sets of molecular retrosynthesis, ReSynZ demonstrates excellent predictive performance compared to existing algorithms. The advantages, such as self-improving model features, flexible reward settings, the potential to surpass human limitations in chemical synthesis route planning, and others, make ReSynZ a valuable tool in chemical synthesis design.

AIGO-DTI: Predicting Drug-Target Interactions Based on Improved Drug Properties Combined with Adaptive Iterative Algorithms.

Artificial intelligence-based methods for predicting drug-target interactions (DTIs) aim to explore reliable drug candidate targets rapidly and cost-effectively to accelerate the drug development process. However, current methods are often limited by the topological regularities of drug molecules, making them difficult to generalize to a broader chemical space. Additionally, the use of similarity to measure DTI network links often introduces noise, leading to false DTI relationships and affecting the prediction accuracy. To address these issues, this study proposes an Adaptive Iterative Graph Optimization (AIGO)-DTI prediction framework. This framework integrates atomic cluster information and enhances molecular features through the design of functional group prompts and graph encoders, optimizing the construction of DTI association networks. Furthermore, the optimization of graph structure is transformed into a node similarity learning problem, utilizing multihead similarity metric functions to iteratively update the network structure to improve the quality of DTI information. Experimental results demonstrate the outstanding performance of AIGO-DTI on multiple public data sets and label reversal data sets. Case studies, molecular docking, and existing research validate its effectiveness and reliability. Overall, the method proposed in this study can construct comprehensive and reliable DTI association network information, providing new graphing and optimization strategies for DTI prediction, which contribute to efficient drug development and reduce target discovery costs.

Enhancing Urban Living Convenience through Plot Patterns: A Quantitative Morphological Study

Living convenience in public service facilities has attracted significant attention as a crucial indicator of urban development and quality improvement. However, the performance of plot patterns—a fundamental unit for precise control when measuring urban quality—influencing living convenience requires in-depth exploration. This study integrates multiple urban datasets with quantitative urban morphology methods to investigate the impact of various plot pattern features on living convenience. Specifically, we investigate the Inner Ring area of Shanghai as an empirical case. The assessment considers the diverse effects of facilities at different distances, accounting for the life radiuses of both older individuals and the general population. Additionally, the analysis of plot patterns includes planar and three-dimensional aspects, controlling key variables such as road network accessibility and centrality. The results indicate that, for small-scale plots, shape has a strong impact, while, for large-scale plots, the division and construction intensity within a block has a stronger influence. Furthermore, plots of different area types have different recommended construction intensities. Notably, for large-scale blocks, it is advisable to maintain a building density of around 0.3. In short, this study contributes to human-centered planning by providing targeted recommendations to address the existing deficiencies in plot morphology regulation and control from the perspective of quantitative urban morphology.

Federated Learning With Only Positive Labels by Exploring Label Correlations.

Federated learning (FL) aims to collaboratively learn a model by using the data from multiple users under privacy constraints. In this article, we study the multilabel classification (MLC) problem under the FL setting, where trivial solution and extremely poor performance may be obtained, especially when only positive data with respect to a single class label is provided for each client. This issue can be addressed by adding a specially designed regularizer on the server side. Although effective sometimes, the label correlations are simply ignored and thus suboptimal performance may be obtained. Besides, it is expensive and unsafe to exchange user's private embeddings between server and clients frequently, especially when training model in the contrastive way. To remedy these drawbacks, we propose a novel and generic method termed federated averaging (FedAvg) by exploring label correlations (FedALCs). Specifically, FedALC estimates the label correlations in the class embedding learning for different label pairs and utilizes it to improve the model training. To further improve the safety and also reduce the communication overhead, we propose a variant to learn fixed class embedding for each client, so that the server and clients only need to exchange class embeddings once. Extensive experiments on multiple popular datasets demonstrate that our FedALC can significantly outperform the existing counterparts.

RFBoost: Understanding and Boosting Deep WiFi Sensing via Physical Data Augmentation

Deep learning shows promising performance in wireless sensing. However, deep wireless sensing (DWS) heavily relies on large datasets. Unfortunately, building comprehensive datasets for DWS is difficult and costly, because wireless data depends on environmental factors and cannot be labeled offline. Despite recent advances in few-shot/cross-domain learning, DWS is still facing data scarcity issues. In this paper, we investigate a distinct perspective of radio data augmentation (RDA) for WiFi sensing and present a data-space solution. Our key insight is that wireless signals inherently exhibit data diversity, contributing more information to be extracted for DWS. We present RFBoost, a simple and effective RDA framework encompassing novel physical data augmentation techniques. We implement RFBoost as a plug-and-play module integrated with existing deep models and evaluate it on multiple datasets. Experimental results demonstrate that RFBoost achieves remarkable average accuracy improvements of 5.4% on existing models without additional data collection or model modifications, and the best-boosted performance outperforms 11 state-of-the-art baseline models without RDA. RFBoost pioneers the study of RDA, an important yet currently underexplored building block for DWS, which we expect to become a standard DWS component of WiFi sensing and beyond. RFBoost is released at https://github.com/aiot-lab/RFBoost.

CrossHAR: Generalizing Cross-dataset Human Activity Recognition via Hierarchical Self-Supervised Pretraining

The increasing availability of low-cost wearable devices and smartphones has significantly advanced the field of sensor-based human activity recognition (HAR), attracting considerable research interest. One of the major challenges in HAR is the domain shift problem in cross-dataset activity recognition, which occurs due to variations in users, device types, and sensor placements between the source dataset and the target dataset. Although domain adaptation methods have shown promise, they typically require access to the target dataset during the training process, which might not be practical in some scenarios. To address these issues, we introduce CrossHAR, a new HAR model designed to improve model performance on unseen target datasets. CrossHAR involves three main steps: (i) CrossHAR explores the sensor data generation principle to diversify the data distribution and augment the raw sensor data. (ii) CrossHAR then employs a hierarchical self-supervised pretraining approach with the augmented data to develop a generalizable representation. (iii) Finally, CrossHAR fine-tunes the pretrained model with a small set of labeled data in the source dataset, enhancing its performance in cross-dataset HAR. Our extensive experiments across multiple real-world HAR datasets demonstrate that CrossHAR outperforms current state-of-the-art methods by 10.83% in accuracy, demonstrating its effectiveness in generalizing to unseen target datasets.