Search

The goal of this work is to develop a task-agnostic feature upsampling operator for dense prediction where the operator is required to facilitate not only region-sensitive tasks like semantic segmentation but also detail-sensitive tasks such as image matting. Prior upsampling operators often can work well in either type of the tasks, but not both. We argue that task-agnostic upsampling should dynamically trade off between semantic preservation and detail delineation, instead of having a bias between the two properties. In this paper, we present FADE, a novel, plug-and-play, lightweight, and task-agnostic upsampling operator by fusing the assets of decoder and encoder features at three levels: i) considering both the encoder and decoder feature in upsampling kernel generation; ii) controlling the per-point contribution of the encoder/decoder feature in upsampling kernels with an efficient semi-shift convolutional operator; and iii) enabling the selective pass of encoder features with a decoder-dependent gating mechanism for compensating details. To improve the practicality of FADE, we additionally study parameter- and memory-efficient implementations of semi-shift convolution. We analyze the upsampling behavior of FADE on toy data and show through large-scale experiments that FADE is task-agnostic with consistent performance improvement on a number of dense prediction tasks with little extra cost. For the first time, we demonstrate robust feature upsampling on both region- and detail-sensitive tasks successfully. Code is made available at: https://github.com/poppinace/fade, Comment: Accepted to International Journal of Computer Vision. Extended version of ECCV 2022 paper at arXiv:2207.10392

High dynamic range (HDR) video rendering from low dynamic range (LDR) videos where frames are of alternate exposure encounters significant challenges, due to the exposure change and absence at each time stamp. The exposure change and absence make existing methods generate flickering HDR results. In this paper, we propose a novel paradigm to render HDR frames via completing the absent exposure information, hence the exposure information is complete and consistent. Our approach involves interpolating neighbor LDR frames in the time dimension to reconstruct LDR frames for the absent exposures. Combining the interpolated and given LDR frames, the complete set of exposure information is available at each time stamp. This benefits the fusing process for HDR results, reducing noise and ghosting artifacts therefore improving temporal consistency. Extensive experimental evaluations on standard benchmarks demonstrate that our method achieves state-of-the-art performance, highlighting the importance of absent exposure completing in HDR video rendering. The code is available at https://github.com/cuijiahao666/NECHDR., Comment: 9 pages, 6 figures, accepted by ACM-MM 2024

We present a time-dependent variational approach with the multiple Davydov $D_2$ trial state to simulate the dynamics of light-matter systems when the field is in a coherent state with an arbitrary finite mean photon number. The variational approach captures not only the system dynamics but also the field dynamics and is applicable to a variety of quantum models of light-matter interaction such as the Jaynes-Cummings model, Rabi model, and Dicke model, and is feasible to tackle the multimode quantized fields. By comparison of the variational and semiclassical dynamics of both the system and field, we illustrate that the variational dynamics from the quantum models agrees with those from the corresponding semiclassical models as long as the mean number of photons is sufficiently large. Moreover, we illustrate that in the crossover between the quantum and semiclassical limits, the quantum corrections lead to the collapse of the oscillations in dynamics, which is absent in the semiclassical models. The variational approach provides a unified treatment of light-matter interaction from the quantum to the semiclassical limit., Comment: 14 pages, 8 figures

Existing WSI analysis methods lie on the consensus that histopathological characteristics of tumors are significant guidance for cancer diagnostics. Particularly, as the evolution of cancers is a continuous process, the correlations and differences across various stages, anatomical locations and patients should be taken into account. However, recent research mainly focuses on the inner-contextual information in a single WSI, ignoring the correlations between slides. To verify whether introducing the slide inter-correlations can bring improvements to WSI representation learning, we propose a generic WSI analysis pipeline SlideGCD that considers the existing multi-instance learning (MIL) methods as the backbone and forge the WSI classification task as a node classification problem. More specifically, SlideGCD declares a node buffer that stores previous slide embeddings for subsequent extensive slide-based graph construction and conducts graph learning to explore the inter-correlations implied in the slide-based graph. Moreover, we frame the MIL classifier and graph learning into two parallel workflows and deploy the knowledge distillation to transfer the differentiable information to the graph neural network. The consistent performance boosting, brought by SlideGCD, of four previous state-of-the-art MIL methods is observed on two TCGA benchmark datasets. The code is available at https://github.com/HFUT-miaLab/SlideGCD., Comment: Accepted for MICCAI 2024

Large-scale pre-training models have promoted the development of histopathology image analysis. However, existing self-supervised methods for histopathology images focus on learning patch features, while there is still a lack of available pre-training models for WSI-level feature learning. In this paper, we propose a novel self-supervised learning framework for pan-cancer WSI-level representation pre-training with the designed position-aware masked autoencoder (PAMA). Meanwhile, we propose the position-aware cross-attention (PACA) module with a kernel reorientation (KRO) strategy and an anchor dropout (AD) mechanism. The KRO strategy can capture the complete semantic structure and eliminate ambiguity in WSIs, and the AD contributes to enhancing the robustness and generalization of the model. We evaluated our method on 6 large-scale datasets from multiple organs for pan-cancer classification tasks. The results have demonstrated the effectiveness of PAMA in generalized and discriminative WSI representation learning and pan-cancer WSI pre-training. The proposed method was also compared with 7 WSI analysis methods. The experimental results have indicated that our proposed PAMA is superior to the state-of-the-art methods.The code and checkpoints are available at https://github.com/WkEEn/PAMA.

Content-based histopathological image retrieval (CBHIR) has gained attention in recent years, offering the capability to return histopathology images that are content-wise similar to the query one from an established database. However, in clinical practice, the continuously expanding size of WSI databases limits the practical application of the current CBHIR methods. In this paper, we propose a Lifelong Whole Slide Retrieval (LWSR) framework to address the challenges of catastrophic forgetting by progressive model updating on continuously growing retrieval database. Our framework aims to achieve the balance between stability and plasticity during continuous learning. To preserve system plasticity, we utilize local memory bank with reservoir sampling method to save instances, which can comprehensively encompass the feature spaces of both old and new tasks. Furthermore, A distance consistency rehearsal (DCR) module is designed to ensure the retrieval queue's consistency for previous tasks, which is regarded as stability within a lifelong CBHIR system. We evaluated the proposed method on four public WSI datasets from TCGA projects. The experimental results have demonstrated the proposed method is effective and is superior to the state-of-the-art methods., Comment: Accepted for MICCAI 2024

Dynamic Neural Radiance Field (NeRF) from monocular videos has recently been explored for space-time novel view synthesis and achieved excellent results. However, defocus blur caused by depth variation often occurs in video capture, compromising the quality of dynamic reconstruction because the lack of sharp details interferes with modeling temporal consistency between input views. To tackle this issue, we propose D2RF, the first dynamic NeRF method designed to restore sharp novel views from defocused monocular videos. We introduce layered Depth-of-Field (DoF) volume rendering to model the defocus blur and reconstruct a sharp NeRF supervised by defocused views. The blur model is inspired by the connection between DoF rendering and volume rendering. The opacity in volume rendering aligns with the layer visibility in DoF rendering.To execute the blurring, we modify the layered blur kernel to the ray-based kernel and employ an optimized sparse kernel to gather the input rays efficiently and render the optimized rays with our layered DoF volume rendering. We synthesize a dataset with defocused dynamic scenes for our task, and extensive experiments on our dataset show that our method outperforms existing approaches in synthesizing all-in-focus novel views from defocus blur while maintaining spatial-temporal consistency in the scene., Comment: Accepted by ECCV 2024

Hybrid non-orthogonal multiple access (NOMA) has recently received significant research interest due to its ability to efficiently use resources from different domains and also its compatibility with various orthogonal multiple access (OMA) based legacy networks. Unlike existing studies on hybrid NOMA that focus on combining NOMA with time-division multiple access (TDMA), this work considers hybrid NOMA assisted orthogonal frequency-division multiple access (OFDMA). In particular, the impact of a unique feature of hybrid NOMA assisted OFDMA, i.e., the availability of users' dynamic channel state information, on the system performance is analyzed from the following two perspectives. From the optimization perspective, analytical results are developed which show that with hybrid NOMA assisted OFDMA, the pure OMA mode is rarely adopted by the users, and the pure NOMA mode could be optimal for minimizing the users' energy consumption, which differs from the hybrid TDMA case. From the statistical perspective, two new performance metrics, namely the power outage probability and the power diversity gain, are developed to quantitatively measure the performance gain of hybrid NOMA over OMA. The developed analytical results also demonstrate the ability of hybrid NOMA to meet the users' diverse energy profiles.

Existing human datasets for avatar creation are typically limited to laboratory environments, wherein high-quality annotations (e.g., SMPL estimation from 3D scans or multi-view images) can be ideally provided. However, their annotating requirements are impractical for real-world images or videos, posing challenges toward real-world applications on current avatar creation methods. To this end, we propose the WildAvatar dataset, a web-scale in-the-wild human avatar creation dataset extracted from YouTube, with $10,000+$ different human subjects and scenes. WildAvatar is at least $10\times$ richer than previous datasets for 3D human avatar creation. We evaluate several state-of-the-art avatar creation methods on our dataset, highlighting the unexplored challenges in real-world applications on avatar creation. We also demonstrate the potential for generalizability of avatar creation methods, when provided with data at scale. We publicly release our data source links and annotations, to push forward 3D human avatar creation and other related fields for real-world applications., Comment: Project page: https://wildavatar.github.io/

Automatic image cropping models predict reframing boxes to enhance image aesthetics. Yet, the scarcity of labeled data hinders the progress of this task. To overcome this limitation, we explore the possibility of utilizing both labeled and unlabeled data together to expand the scale of training data for image cropping models. This idea can be implemented in a pseudo-labeling way: producing pseudo labels for unlabeled data by a teacher model and training a student model with these pseudo labels. However, the student may learn from teacher's mistakes. To address this issue, we propose the multi-policy viewfinder network (MPV-Net) that offers diverse refining policies to rectify the mistakes in original pseudo labels from the teacher. The most reliable policy is selected to generate trusted pseudo labels. The reliability of policies is evaluated via the robustness against box jittering. The efficacy of our method can be evaluated by the improvement compared to the supervised baseline which only uses labeled data. Notably, our MPV-Net outperforms off-the-shelf pseudo-labeling methods, yielding the most substantial improvement over the supervised baseline. Furthermore, our approach achieves state-of-the-art results on both the FCDB and FLMS datasets, signifying the superiority of our approach., Comment: 18 pages, 8figures

Despite the rapid development of Chinese vision-language models (VLMs), most existing Chinese vision-language (VL) datasets are constructed on Western-centric images from existing English VL datasets. The cultural bias in the images makes these datasets unsuitable for evaluating VLMs in Chinese culture. To remedy this issue, we present a new Chinese Vision- Language Understanding Evaluation (CVLUE) benchmark dataset, where the selection of object categories and images is entirely driven by Chinese native speakers, ensuring that the source images are representative of Chinese culture. The benchmark contains four distinct VL tasks ranging from image-text retrieval to visual question answering, visual grounding and visual dialogue. We present a detailed statistical analysis of CVLUE and provide a baseline performance analysis with several open-source multilingual VLMs on CVLUE and its English counterparts to reveal their performance gap between English and Chinese. Our in-depth category-level analysis reveals a lack of Chinese cultural knowledge in existing VLMs. We also find that fine-tuning on Chinese culture-related VL datasets effectively enhances VLMs' understanding of Chinese culture.

Despite its popularity in sentence-level relation extraction, distantly supervised data is rarely utilized by existing work in document-level relation extraction due to its noisy nature and low information density. Among its current applications, distantly supervised data is mostly used as a whole for pertaining, which is of low time efficiency. To fill in the gap of efficient and robust utilization of distantly supervised training data, we propose Efficient Multi-Supervision for document-level relation extraction, in which we first select a subset of informative documents from the massive dataset by combining distant supervision with expert supervision, then train the model with Multi-Supervision Ranking Loss that integrates the knowledge from multiple sources of supervision to alleviate the effects of noise. The experiments demonstrate the effectiveness of our method in improving the model performance with higher time efficiency than existing baselines.

Time-Sensitive Networking (TSN) serves as a one-size-fits-all solution for mixed-criticality communication, in which flow scheduling is vital to guarantee real-time transmissions. Traditional approaches statically assign priorities to flows based on their associated applications, resulting in significant queuing delays. In this paper, we observe that assigning different priorities to a flow leads to varying delays due to different shaping mechanisms applied to different flow types. Leveraging this insight, we introduce a new scheduling method in mixed-criticality TSN that incorporates a priority adjustment scheme among diverse flow types to mitigate queuing delays and enhance schedulability. Specifically, we propose dependency-aware priority adjustment algorithms tailored to different link-overlapping conditions. Experiments in various settings validate the effectiveness of the proposed method, which enhances the schedulability by 20.57% compared with the SOTA method., Comment: This paper has been accepted by IWQoS'24

Large-scale datasets with point-wise semantic and instance labels are crucial to 3D instance segmentation but also expensive. To leverage unlabeled data, previous semi-supervised 3D instance segmentation approaches have explored self-training frameworks, which rely on high-quality pseudo labels for consistency regularization. They intuitively utilize both instance and semantic pseudo labels in a joint learning manner. However, semantic pseudo labels contain numerous noise derived from the imbalanced category distribution and natural confusion of similar but distinct categories, which leads to severe collapses in self-training. Motivated by the observation that 3D instances are non-overlapping and spatially separable, we ask whether we can solely rely on instance consistency regularization for improved semi-supervised segmentation. To this end, we propose a novel self-training network InsTeacher3D to explore and exploit pure instance knowledge from unlabeled data. We first build a parallel base 3D instance segmentation model DKNet, which distinguishes each instance from the others via discriminative instance kernels without reliance on semantic segmentation. Based on DKNet, we further design a novel instance consistency regularization framework to generate and leverage high-quality instance pseudo labels. Experimental results on multiple large-scale datasets show that the InsTeacher3D significantly outperforms prior state-of-the-art semi-supervised approaches. Code is available: https://github.com/W1zheng/InsTeacher3D., Comment: 14 pages, 10 figures

The \emph{matching book thickness} $mbt(G)$ of $G$ is the minimum integer $m$ such that an $m$-page matching book embedding exists. A graph $G$ is called \emph{dispersable} if $mbt(G)=\Delta(G)$, \emph{nearly dispersable} if $mbt(G)=\Delta(G)+1$. Recently, the authors determined the nearly dispersability of odd toroidal grids $T_{s,t}$. In this note, we further present a brief proof for this result.

With built-in eye-tracking cameras, the Apple Vision Pro (AVP) enables gaze-based interaction, eye image rendering on external screens, and iris recognition for device unlocking. One of the technological advancements of the AVP is its heavy reliance on gaze- and gesture-based interaction. However, limited information is available regarding the specifics of the eye-tracking device in the AVP, and raw gaze data is inaccessible to developers. This study evaluated the eye-tracking accuracy of the AVP, leveraging foveated rendering, and examined how tracking accuracy relates to user-reported usability. The results revealed an overall gaze error of 2.5{\deg} (or 61.95 pixels) within a tested field of view (FOV) of approximately 34{\deg} x 18{\deg}. As expected, the lowest gaze error was observed in the central FOV, with higher gaze errors in peripheral areas. The usability and learnability scores of the AVP, measured using the standard System Usability Scale (SUS), were 73 and 70, respectively. Importantly, no statistically reliable correlation between gaze error and usability scores was found. These results suggest that the eye-tracking accuracy of the AVP is comparable to other VR/AR headsets. While eye-tracking accuracy is critical for gaze-based interaction, it is not the sole determinant of user experience in AR/VR., Comment: 10 pages, 7 figures and 2 tables

In this paper, the dispersability of the Cartesian graph bundle over two cycles is completely solved. We show the Cartesian graph bundle $G$ over two cycles is dispersable if $G$ is bipartite; otherwise, $G$ is nearly dispersable., Comment: arXiv admin note: text overlap with arXiv:2310.06612

This paper investigates federated learning in a wireless communication system, where random device selection is employed with non-independent and identically distributed (non-IID) data. The analysis indicates that while training deep learning networks using federated stochastic gradient descent (FedSGD) on non-IID datasets, device selection can generate gradient errors that accumulate, leading to potential weight divergence. To mitigate training divergence, we design an age-weighted FedSGD to scale local gradients according to the previous state of devices. To further improve learning performance by increasing device participation under the maximum time consumption constraint, we formulate an energy consumption minimization problem by including resource allocation and sub-channel assignment. By transforming the resource allocation problem into convex and utilizing KKT conditions, we derived the optimal resource allocation solution. Moreover, this paper develops a matching based algorithm to generate the enhanced sub-channel assignment. Simulation results indicate that i) age-weighted FedSGD is able to outperform conventional FedSGD in terms of convergence rate and achievable accuracy, and ii) the proposed resource allocation and sub-channel assignment strategies can significantly reduce energy consumption and improve learning performance by increasing the number of selected devices.

We present MVSGaussian, a new generalizable 3D Gaussian representation approach derived from Multi-View Stereo (MVS) that can efficiently reconstruct unseen scenes. Specifically, 1) we leverage MVS to encode geometry-aware Gaussian representations and decode them into Gaussian parameters. 2) To further enhance performance, we propose a hybrid Gaussian rendering that integrates an efficient volume rendering design for novel view synthesis. 3) To support fast fine-tuning for specific scenes, we introduce a multi-view geometric consistent aggregation strategy to effectively aggregate the point clouds generated by the generalizable model, serving as the initialization for per-scene optimization. Compared with previous generalizable NeRF-based methods, which typically require minutes of fine-tuning and seconds of rendering per image, MVSGaussian achieves real-time rendering with better synthesis quality for each scene. Compared with the vanilla 3D-GS, MVSGaussian achieves better view synthesis with less training computational cost. Extensive experiments on DTU, Real Forward-facing, NeRF Synthetic, and Tanks and Temples datasets validate that MVSGaussian attains state-of-the-art performance with convincing generalizability, real-time rendering speed, and fast per-scene optimization., Comment: ECCV2024, Project page: https://mvsgaussian.github.io/ , Code: https://github.com/TQTQliu/MVSGaussian

A fundamental objective of the forthcoming sixth-generation wireless networks is to concurrently serve a vast array of devices many of which, such as Internet-of-Things (IoT) sensors, are projected to have low power requirements or even operate in a battery-free manner. To achieve this goal, non-orthogonal multiple access (NOMA) and ambient backscatter communications (AmBC) are regarded as two pivotal and promising technologies. In this work, we present a novel analytical framework for studying the reliability and security of uplink NOMA-based AmBC systems. Specifically, closed-form analytical expressions for both NOMA-users' and IoT backscatter device's (BD's) outage probabilities (OPs) are derived for both cases of perfect and imperfect successive interference cancellation (SIC). In addition, assuming that one NOMA-user transmits an artificial noise in order to enhance system's security, the physical layer security (PLS) of the system is investigated by extracting analytical expressions for NOMA-users' and BD's intercept probabilities (IPs). To gain insightful understandings, an asymptotic analysis is carried out by focusing on the high signal-to-noise (SNR) regime, which reveals that NOMA-users and BDs face outage floors in the high SNR regime as well as that IPs reach constant values at high SNR. Additionally, practical insights regarding how different system parameters affect these OP floors and IP constant values are extracted. Numerical results verify the accuracy of othe developed theoretical framework, offer performance comparisons between the presented NOMA-based AmBC system and a conventional orthogonal multiple access-based AmBC system, and reveal the impact of different system parameters on the reliability and security of NOMA-based AmBC networks., Comment: 14 pages, 6 figures

Despite their exceptional performance on various tasks after fine-tuning, pre-trained language models (PLMs) face significant challenges due to growing privacy concerns with data in centralized training methods. We consider federated learning (FL) to fine-tune PLMs in this paper. However, the substantial number of parameters in PLMs poses significant difficulties for client devices with limited communication and computational resources. One promising solution is to exploit parameter-efficient fine-tuning (PEFT) into FL, which trains a much smaller set of parameters than full parameter fine-tuning (FFT). Although remarkably improving training efficiency, PEFT methods may lead to degraded performance especially when data across different clients are non i.i.d, as revealed by experimental results. To overcome this, we propose FeDeRA, which extends and improves a widely used PEFT method, i.e., low-rank adaption (LoRA). FeDeRA follows LoRA by decomposing the weight matrices of the PLMs into low-rank matrices, which allows for more efficient computation and parameter updates during fine-tuning. Different from LoRA which simply initializes these low-rank matrices by random sampling or zeros, the proposed FeDeRA initializes these matrices by the results of performing singular value decomposition (SVD) on the pre-trained weight matrices. Extensive experiments across various tasks and datasets show that FeDeRA outperforms the considered PEFT baselines and is comparable to or even surpasses FFT method within the FL setting in terms of task performance. Moreover, FeDeRA requires only 1% trainable paramentes compared to FFT, significantly reducing training time costs by more than 90% to achieve the same task performance level. The experimental results also highlight the robustness of FeDeRA against data heterogeneity, as it maintains stable task performance even as data heterogeneity increases.

Cross-center data heterogeneity and annotation unreliability significantly challenge the intelligent diagnosis of diseases using brain signals. A notable example is the EEG-based diagnosis of neurodegenerative diseases, which features subtler abnormal neural dynamics typically observed in small-group settings. To advance this area, in this work, we introduce a transferable framework employing Manifold Attention and Confidence Stratification (MACS) to diagnose neurodegenerative disorders based on EEG signals sourced from four centers with unreliable annotations. The MACS framework's effectiveness stems from these features: 1) The Augmentor generates various EEG-represented brain variants to enrich the data space; 2) The Switcher enhances the feature space for trusted samples and reduces overfitting on incorrectly labeled samples; 3) The Encoder uses the Riemannian manifold and Euclidean metrics to capture spatiotemporal variations and dynamic synchronization in EEG; 4) The Projector, equipped with dual heads, monitors consistency across multiple brain variants and ensures diagnostic accuracy; 5) The Stratifier adaptively stratifies learned samples by confidence levels throughout the training process; 6) Forward and backpropagation in MACS are constrained by confidence stratification to stabilize the learning system amid unreliable annotations. Our subject-independent experiments, conducted on both neurocognitive and movement disorders using cross-center corpora, have demonstrated superior performance compared to existing related algorithms. This work not only improves EEG-based diagnostics for cross-center and small-setting brain diseases but also offers insights into extending MACS techniques to other data analyses, tackling data heterogeneity and annotation unreliability in multimedia and multimodal content understanding.

Generalizable NeRF aims to synthesize novel views for unseen scenes. Common practices involve constructing variance-based cost volumes for geometry reconstruction and encoding 3D descriptors for decoding novel views. However, existing methods show limited generalization ability in challenging conditions due to inaccurate geometry, sub-optimal descriptors, and decoding strategies. We address these issues point by point. First, we find the variance-based cost volume exhibits failure patterns as the features of pixels corresponding to the same point can be inconsistent across different views due to occlusions or reflections. We introduce an Adaptive Cost Aggregation (ACA) approach to amplify the contribution of consistent pixel pairs and suppress inconsistent ones. Unlike previous methods that solely fuse 2D features into descriptors, our approach introduces a Spatial-View Aggregator (SVA) to incorporate 3D context into descriptors through spatial and inter-view interaction. When decoding the descriptors, we observe the two existing decoding strategies excel in different areas, which are complementary. A Consistency-Aware Fusion (CAF) strategy is proposed to leverage the advantages of both. We incorporate the above ACA, SVA, and CAF into a coarse-to-fine framework, termed Geometry-aware Reconstruction and Fusion-refined Rendering (GeFu). GeFu attains state-of-the-art performance across multiple datasets. Code is available at https://github.com/TQTQliu/GeFu ., Comment: Accepted by CVPR 2024. Project page: https://gefucvpr24.github.io

Large language models (LLMs) have shown remarkable performance on a variety of NLP tasks, and are being rapidly adopted in a wide range of use cases. It is therefore of vital importance to holistically evaluate the factuality of their generated outputs, as hallucinations remain a challenging issue. In this work, we focus on assessing LLMs' ability to recall factual knowledge learned from pretraining, and the factors that affect this ability. To that end, we construct FACT-BENCH, a representative benchmark covering 20 domains, 134 property types, 3 answer types, and different knowledge popularity levels. We benchmark 31 models from 10 model families and provide a holistic assessment of their strengths and weaknesses. We observe that instruction-tuning hurts knowledge recall, as pretraining-only models consistently outperform their instruction-tuned counterparts, and positive effects of model scaling, as larger models outperform smaller ones for all model families. However, the best performance from GPT-4 still represents a large gap with the upper-bound. We additionally study the role of in-context exemplars using counterfactual demonstrations, which lead to significant degradation of factual knowledge recall for large models. By further decoupling model known and unknown knowledge, we find the degradation is attributed to exemplars that contradict a model's known knowledge, as well as the number of such exemplars. Lastly, we fine-tune LLaMA-7B in different settings of known and unknown knowledge. In particular, fine-tuning on a model's known knowledge is beneficial, and consistently outperforms fine-tuning on unknown and mixed knowledge. We will make our benchmark publicly available.

In this paper, we present RStab, a novel framework for video stabilization that integrates 3D multi-frame fusion through volume rendering. Departing from conventional methods, we introduce a 3D multi-frame perspective to generate stabilized images, addressing the challenge of full-frame generation while preserving structure. The core of our approach lies in Stabilized Rendering (SR), a volume rendering module, which extends beyond the image fusion by incorporating feature fusion. The core of our RStab framework lies in Stabilized Rendering (SR), a volume rendering module, fusing multi-frame information in 3D space. Specifically, SR involves warping features and colors from multiple frames by projection, fusing them into descriptors to render the stabilized image. However, the precision of warped information depends on the projection accuracy, a factor significantly influenced by dynamic regions. In response, we introduce the Adaptive Ray Range (ARR) module to integrate depth priors, adaptively defining the sampling range for the projection process. Additionally, we propose Color Correction (CC) assisting geometric constraints with optical flow for accurate color aggregation. Thanks to the three modules, our RStab demonstrates superior performance compared with previous stabilizers in the field of view (FOV), image quality, and video stability across various datasets., Comment: Accepted by CVPR 2024

Physics-based models are computationally time-consuming and infeasible for real-time scenarios of urban drainage networks, and a surrogate model is needed to accelerate the online predictive modelling. Fully-connected neural networks (NNs) are potential surrogate models, but may suffer from low interpretability and efficiency in fitting complex targets. Owing to the state-of-the-art modelling power of graph neural networks (GNNs) and their match with urban drainage networks in the graph structure, this work proposes a GNN-based surrogate of the flow routing model for the hydraulic prediction problem of drainage networks, which regards recent hydraulic states as initial conditions, and future runoff and control policy as boundary conditions. To incorporate hydraulic constraints and physical relationships into drainage modelling, physics-guided mechanisms are designed on top of the surrogate model to restrict the prediction variables with flow balance and flooding occurrence constraints. According to case results in a stormwater network, the GNN-based model is more cost-effective with better hydraulic prediction accuracy than the NN-based model after equal training epochs, and the designed mechanisms further limit prediction errors with interpretable domain knowledge. As the model structure adheres to the flow routing mechanisms and hydraulic constraints in urban drainage networks, it provides an interpretable and effective solution for data-driven surrogate modelling. Simultaneously, the surrogate model accelerates the predictive modelling of urban drainage networks for real-time use compared with the physics-based model.

We investigate the ground-state phase diagram of a binary mixture of Bose-Einstein condensates (BECs) with competing interspecies $s$- and $p$-wave interactions. Exploiting a pseudopotential model for the $l=1$ partial wave, we derive an extended Gross-Pitaevskii (GP) equation for the BEC mixture that incorporates both $s$- and $p$-wave interactions. Based on it, we study the miscible-immiscible transition of a binary BEC mixture in the presence of interspecies $p$-wave interaction, by combining numerical solution of the GP equation and Gaussian variational analysis. Our study uncovers a dual effect -- either enhance or reduce miscibility -- of positive interspecies $p$-wave interaction, which can be precisely controlled by adjusting relevant experimental parameters. By complete characterizing the miscibility phase diagram, we establish a promising avenue towards experimental control of the miscibility of binary BEC mixtures via high partial-wave interactions., Comment: 10+3 pages, 6 figures, Phys. Rev. A (2024)

Bedload transport occurs when a bed composed of sedimentary grains becomes mobile in response to the shearing by a flow of liquid. It shapes the landscapes of Earth and other planetary bodies by promoting the formation and growth of various multiscale geological features. Estimating the rate at which such processes take place requires accurate bedload flux predictions. However, even for highly idealized conditions in the laboratory, study-to-study variability of reported bedload flux measurements borders an order of magnitude. This uncertainty stems from physically poorly supported, typically empirical methods of determining the transport-driving bed shear stress, especially for very narrow or shallow channel flows, and from study-to-study grain shape variations. Here, we derive a non-empirical method of bed shear stress determination and apply it to a number of independent grain-shape-controlled data sets, based on well-controlled experiments and CFD-DEM simulations, for a very diverse range of transport conditions. An existing physical bedload model, here generalized to account for grain shape variability, predicts almost all these data within a factor of 1.3, whereas a prominent alternative model (Deal et al., Nature 613, 298-302, 2023) seems falsified.

Multiple access techniques are fundamental to the design of wireless communication systems, since many crucial components of such systems depend on the choice of the multiple access technique. Because of the importance of multiple access, there has been an ongoing quest during the past decade to develop next generation multiple access (NGMA). Among those potential candidates for NGMA, non-orthogonal multiple access (NOMA) has received significant attention from both the industrial and academic research communities, and has been highlighted in the recently published International Mobile Telecommunications (IMT)-2030 Framework. However, there is still no consensus in the research community about how exactly NOMA assisted NGMA should be designed. This perspective is to outline three important features of NOMA assisted NGMA, namely multi-domain utilization, multi-mode compatibility, and multi-dimensional optimality, where important directions for future research into the design of NOMA assisted NGMA are also discussed.

While existing studies have highlighted the advantages of deep learning (DL)-based joint source-channel coding (JSCC) schemes in enhancing transmission efficiency, they often overlook the crucial aspect of resource management during the deployment phase. In this paper, we propose an approach to minimize the transmission latency in an uplink JSCC-based system. We first analyze the correlation between end-to-end latency and task performance, based on which the end-to-end delay model for each device is established. Then, we formulate a non-convex optimization problem aiming at minimizing the maximum end-to-end latency across all devices, which is proved to be NP-hard. We then transform the original problem into a more tractable one, from which we derive the closed form solution on the optimal compression ratio, truncation threshold selection policy, and resource allocation strategy. We further introduce a heuristic algorithm with low complexity, leveraging insights from the structure of the optimal solution. Simulation results demonstrate that both the proposed optimal algorithm and the heuristic algorithm significantly reduce end-to-end latency. Notably, the proposed heuristic algorithm achieves nearly the same performance to the optimal solution but with considerably lower computational complexity., Comment: 7 Pages, 5 Figures, accepted by 2024 IEEE ICC Workshop

In this paper, we propose a multi-input multi-output transmit beamforming optimization model for joint radar sensing and multi-user communications, where the design of the beamformers is formulated as an optimization problem whose objective is a weighted combination of the sum rate and the Cram\'{e}r-Rao bound, subject to the transmit power budget. Obtaining the global solution for the formulated nonconvex problem is a challenging task, since the sum-rate maximization problem itself (even without considering the sensing metric) is known to be NP-hard. The main contributions of this paper are threefold. Firstly, we derive an optimal closed-form solution to the formulated problem in the single-user case and the multi-user case where the channel vectors of different users are orthogonal. Secondly, for the general multi-user case, we propose a novel branch and bound (B\&B) algorithm based on the McCormick envelope relaxation. The proposed algorithm is guaranteed to find the globally optimal solution to the formulated problem. Thirdly, we design a graph neural network (GNN) based pruning policy to determine irrelevant nodes that can be directly pruned in the proposed B\&B algorithm, thereby significantly reducing the number of unnecessary enumerations therein and improving its computational efficiency. Simulation results show the efficiency of the proposed vanilla and GNN-based accelerated B\&B algorithms., Comment: Submitted for possible publication

Hybrid non-orthogonal multiple access (H-NOMA) has recently received significant attention as a general framework of multiple access, where both conventional orthogonal multiple access (OMA) and pure NOMA are its special cases. This paper focuses on the application of H-NOMA to ambient Internet of Things (IoT) with energy-constrained devices, where a new backscatter communication (BackCom) assisted H-NOMA uplink scheme is developed. Resource allocation for H-NOMA uplink transmission is also considered, where an overall power minimization problem is formulated. Insightful understandings for the key features of BackCom assisted H-NOMA and its difference from conventional H-NOMA are illustrated by developing analytical results for the two-user special case. For the general multi-user scenario, two algorithms, one based on the branch-bound (BB) principle and the other based on successive convex approximation (SCA), are developed to realize different tradeoffs between the system performance and complexity. The numerical results are also provided to verify the accuracy of the developed analytical results and demonstrate the performance gain of H-NOMA over OMA.

We introduce in-context matting, a novel task setting of image matting. Given a reference image of a certain foreground and guided priors such as points, scribbles, and masks, in-context matting enables automatic alpha estimation on a batch of target images of the same foreground category, without additional auxiliary input. This setting marries good performance in auxiliary input-based matting and ease of use in automatic matting, which finds a good trade-off between customization and automation. To overcome the key challenge of accurate foreground matching, we introduce IconMatting, an in-context matting model built upon a pre-trained text-to-image diffusion model. Conditioned on inter- and intra-similarity matching, IconMatting can make full use of reference context to generate accurate target alpha mattes. To benchmark the task, we also introduce a novel testing dataset ICM-$57$, covering 57 groups of real-world images. Quantitative and qualitative results on the ICM-57 testing set show that IconMatting rivals the accuracy of trimap-based matting while retaining the automation level akin to automatic matting. Code is available at https://github.com/tiny-smart/in-context-matting, Comment: Accepted to CVPR 2024. Code is available at https://github.com/tiny-smart/in-context-matting

This paper addresses the problem of group target tracking (GTT), wherein multiple closely spaced targets within a group pose a coordinated motion. To improve the tracking performance, the labeled random finite sets (LRFSs) theory is adopted, and this paper develops a new kind of LRFSs, i.e., augmented LRFSs, which introduces group information into the definition of LRFSs. Specifically, for each element in an LRFS, the kinetic states, track label, and the corresponding group information of its represented target are incorporated. Furthermore, by means of the labeled multi-Bernoulli (LMB) filter with the proposed augmented LRFSs, the group structure is iteratively propagated and updated during the tracking process, which achieves the simultaneously estimation of the kinetic states, track label, and the corresponding group information of multiple group targets, and further improves the GTT tracking performance. Finally, simulation experiments are provided, which well demonstrates the effectiveness of the labeled multi-Bernoulli filter with the proposed augmented LRFSs for GTT tracking.

The perception of motion behavior in a dynamic environment holds significant importance for autonomous driving systems, wherein class-agnostic motion prediction methods directly predict the motion of the entire point cloud. While most existing methods rely on fully-supervised learning, the manual labeling of point cloud data is laborious and time-consuming. Therefore, several annotation-efficient methods have been proposed to address this challenge. Although effective, these methods rely on weak annotations or additional multi-modal data like images, and the potential benefits inherent in the point cloud sequence are still underexplored. To this end, we explore the feasibility of self-supervised motion prediction with only unlabeled LiDAR point clouds. Initially, we employ an optimal transport solver to establish coarse correspondences between current and future point clouds as the coarse pseudo motion labels. Training models directly using such coarse labels leads to noticeable spatial and temporal prediction inconsistencies. To mitigate these issues, we introduce three simple spatial and temporal regularization losses, which facilitate the self-supervised training process effectively. Experimental results demonstrate the significant superiority of our approach over the state-of-the-art self-supervised methods., Comment: Accepted by CVPR2024

As a revolutionary technology, reconfigurable intelligent surface (RIS) has been deemed as an indispensable part of the 6th generation communications due to its inherent ability to regulate the wireless channels. However, passive RIS (PRIS) still suffers from some pressing issues, one of which is that the fading of the entire reflection link is proportional to the product of the distances from the base station to the PRIS and from the PRIS to the users, i.e., the productive attenuation. To tackle this problem, active RIS (ARIS) has been proposed to reconfigure the wireless propagation condition and alleviate the productive attenuation. In this paper, we investigate the physical layer security of the ARIS assisted non-orthogonal multiple access (NOMA) networks with the attendance of external and internal eavesdroppers. To be specific, the closed-form expressions of secrecy outage probability (SOP) and secrecy system throughput are derived by invoking both imperfect successive interference cancellation (ipSIC) and perfect SIC. The secrecy diversity orders of legitimate users are obtained at high signal-to-noise ratios. Numerical results are presented to verify the accuracy of the theoretical expressions and indicate that: i) The SOP of ARIS assisted NOMA networks exceeds that of PRIS-NOMA, ARIS/PRIS-assisted orthogonal multiple access (OMA); ii) Due to the balance between the thermal noise and residual interference, introducing excess reconfigurable elements at ARIS is not helpful to reduce the SOP; and iii) The secrecy throughput performance of ARIS-NOMA networks outperforms that of PRIS-NOMA and ARIS/PRIS-OMA networks., Comment: This paper has been accepted for publication by IEEE Transactions on Wireless Communications

Reconfigurable intelligent surface (RIS) has been regarded as a promising technology since it has ability to create the favorable channel conditions. This paper investigates the secure communications of RIS assisted non-orthogonal multiple access (NOMA) networks, where both external and internal eavesdropping scenarios are taken into consideration. More specifically, novel approximate and asymptotic expressions of secrecy outage probability (SOP) for the k-th legitimate user (LU) are derived by invoking imperfect successive interference cancellation (ipSIC) and perfect successive interference cancellation (pSIC). To characterize the secrecy performance of RIS-NOMA networks, the diversity order of the k-th LU with ipSIC/pSIC is obtained in the high signal-to-noise ratio region. The secrecy system throughput of RIS-NOMA networks is discussed in delay-limited transmission mode. Numerical results are presented to verify theoretical analysis that: i) The SOP of RIS-NOMA networks is superior to that of RIS assisted orthogonal multiple access (OMA) and conventional cooperative communication schemes; ii) As the number of reflecting elements increases, the RIS-NOMA networks are capable of achieving the enhanced secrecy performance; and iii) The RIS-NOMA networks have better secrecy system throughput than that of RIS-OMA networks and conventional cooperative communication schemes., Comment: This paper has been published in IEEE Transactions on Vehicular Technology

Recent advancements in dynamic neural radiance field methods have yielded remarkable outcomes. However, these approaches rely on the assumption of sharp input images. When faced with motion blur, existing dynamic NeRF methods often struggle to generate high-quality novel views. In this paper, we propose DyBluRF, a dynamic radiance field approach that synthesizes sharp novel views from a monocular video affected by motion blur. To account for motion blur in input images, we simultaneously capture the camera trajectory and object Discrete Cosine Transform (DCT) trajectories within the scene. Additionally, we employ a global cross-time rendering approach to ensure consistent temporal coherence across the entire scene. We curate a dataset comprising diverse dynamic scenes that are specifically tailored for our task. Experimental results on our dataset demonstrate that our method outperforms existing approaches in generating sharp novel views from motion-blurred inputs while maintaining spatial-temporal consistency of the scene., Comment: Accepted by CVPR 2024. Project page: https://huiqiang-sun.github.io/dyblurf/

Most existing one-shot skeleton-based action recognition focuses on raw low-level information (e.g., joint location), and may suffer from local information loss and low generalization ability. To alleviate these, we propose to leverage text description generated from large language models (LLM) that contain high-level human knowledge, to guide feature learning, in a global-local-global way. Particularly, during training, we design $2$ prompts to gain global and local text descriptions of each action from an LLM. We first utilize the global text description to guide the skeleton encoder focus on informative joints (i.e.,global-to-local). Then we build non-local interaction between local text and joint features, to form the final global representation (i.e., local-to-global). To mitigate the asymmetry issue between the training and inference phases, we further design a dual-branch architecture that allows the model to perform novel class inference without any text input, also making the additional inference cost neglectable compared with the base skeleton encoder. Extensive experiments on three different benchmarks show that CrossGLG consistently outperforms the existing SOTA methods with large margins, and the inference cost (model size) is only $2.8$\% than the previous SOTA. CrossGLG can also serve as a plug-and-play module that can substantially enhance the performance of different SOTA skeleton encoders with a neglectable cost during inference. The source code will be released soon.

This paper establishes an extreme $C^k$ reducibility theorem of quasi-periodic $SL(2, \mathbb{R})$ cocycles in the local perturbative region, revealing both the essence of Eliasson [Commun.Math.Phys.1992] and Hou-You [Invent.Math.2012] in respectively the non-resonant and resonant cases. By paralleling further the reducibility process with the almost reducibility, we are able to acquire the least initial regularity as well as the least loss of regularity for the whole KAM iterations. This, in return, makes various spectral applications of quasi-periodic Schr\"odinger operators wide open.

Poincar\'{e}'s classical results [H. Poincar\'{e}, Sur l'int\'{e}gration des \'{e}quations diff\'{e}rentielles du premier order et du premier degr\'{e} I and II, Rend. Circ. Mat. Palermo 5 (1891) 161-191; 11 (1897) 193-239] first provide a link between the existence of analytic first integrals and the resonant relations for analytic dynamical systems. In this paper, we show that by appropriately selecting the definition of the stochastic local first integrals, we are able to obtain the stochastic version of Poincar\'e non-integrability theorem. More specifically, we introduce two definitions of local first integrals for stochastic differential equations (SDEs) in the sense of probability one and expectation, respectively. We present the necessary conditions for the existence of functionally independent analytic or rational first integrals of SDEs via the resonances. We also show that for given integrable ordinary differential equations with some nondegeneracy conditions, there exists a linear stochastic perturbation such that the corresponding disturbed SDEs have no any analytic first integrals. Some examples are given to illustrate our results.

Current 3D stylization methods often assume static scenes, which violates the dynamic nature of our real world. To address this limitation, we present S-DyRF, a reference-based spatio-temporal stylization method for dynamic neural radiance fields. However, stylizing dynamic 3D scenes is inherently challenging due to the limited availability of stylized reference images along the temporal axis. Our key insight lies in introducing additional temporal cues besides the provided reference. To this end, we generate temporal pseudo-references from the given stylized reference. These pseudo-references facilitate the propagation of style information from the reference to the entire dynamic 3D scene. For coarse style transfer, we enforce novel views and times to mimic the style details present in pseudo-references at the feature level. To preserve high-frequency details, we create a collection of stylized temporal pseudo-rays from temporal pseudo-references. These pseudo-rays serve as detailed and explicit stylization guidance for achieving fine style transfer. Experiments on both synthetic and real-world datasets demonstrate that our method yields plausible stylized results of space-time view synthesis on dynamic 3D scenes., Comment: Accepted by CVPR 2024. Project page: https://xingyi-li.github.io/s-dyrf/

Recently, promptable segmentation models, such as the Segment Anything Model (SAM), have demonstrated robust zero-shot generalization capabilities on static images. These promptable models exhibit denoising abilities for imprecise prompt inputs, such as imprecise bounding boxes. In this paper, we explore the potential of applying SAM to track and segment objects in videos where we recognize the tracking task as a prompt denoising task. Specifically, we iteratively propagate the bounding box of each object's mask in the preceding frame as the prompt for the next frame. Furthermore, to enhance SAM's denoising capability against position and size variations, we propose a multi-prompt strategy where we provide multiple jittered and scaled box prompts for each object and preserve the mask prediction with the highest semantic similarity to the template mask. We also introduce a point-based refinement stage to handle occlusions and reduce cumulative errors. Without involving tracking modules, our approach demonstrates comparable performance in video object/instance segmentation tasks on three datasets: DAVIS2017, YouTubeVOS2018, and UVO, serving as a concise baseline and endowing SAM-based downstream applications with tracking capabilities.

This study explores the benefits of integrating the novel clustered federated learning (CFL) approach with non-orthogonal multiple access (NOMA) under non-independent and identically distributed (non-IID) datasets, where multiple devices participate in the aggregation with time limitations and a finite number of sub-channels. A detailed theoretical analysis of the generalization gap that measures the degree of non-IID in the data distribution is presented. Following that, solutions to address the challenges posed by non-IID conditions are proposed with the analysis of the properties. Specifically, users' data distributions are parameterized as concentration parameters and grouped using spectral clustering, with Dirichlet distribution serving as the prior. The investigation into the generalization gap and convergence rate guides the design of sub-channel assignments through the matching-based algorithm, and the power allocation is achieved by Karush-Kuhn-Tucker (KKT) conditions with the derived closed-form solution. The extensive simulation results show that the proposed cluster-based FL framework can outperform FL baselines in terms of both test accuracy and convergence rate. Moreover, jointly optimizing sub-channel and power allocation in NOMA-enhanced networks can lead to a significant improvement.

Existing explanation results of functional connectivity (FC) are normally generated by using classification result labels and correlation analysis methods such as Pearson's correlation or gradient backward. However, the diagnostic model is still trained on the black box model and might lack the attention of FCs in important regions during the training. To enhance the explainability and improve diagnostic performance, providing prior knowledge on neurodegeneration-related regions when healthy subjects (HC) develop into subject cognitive decline (SCD) and mild cognitive impairment (MCI) for the diagnostic model is a key step. To better determine the neurodegeneration-related regions, we employ counterfactual reasoning to generate the target label FC matrices derived from source label FC and then subtract source label FC with target label FC. The counterfactual reasoning architecture is constructed by adaptive forward and backward transformer generative adversarial network (AFBT GAN), which is specifically designed by network property in FC and inverse patch embedding operation in the transformer. The specific design can make the model focus more on the current network correlation and employ the global insight of the transformer to reconstruct FC, which both help the generation of high-quality target label FC. The validation experiments are conducted on both clinical and public datasets, the generated attention map are both vital correlated to cognitive function and the diagnostic performance is also significant. The code is available at https://github.com/SXR3015/AFBT-GAN., Comment: 10 pages, 5 figures

In this paper, we propose a unified framework of inexact stochastic Alternating Direction Method of Multipliers (ADMM) for solving nonconvex problems subject to linear constraints, whose objective comprises an average of finite-sum smooth functions and a nonsmooth but possibly nonconvex function. The new framework is highly versatile. Firstly, it not only covers several existing algorithms such as SADMM, SVRG-ADMM, and SPIDER-ADMM but also guides us to design a novel accelerated hybrid stochastic ADMM algorithm, which utilizes a new hybrid estimator to trade-off variance and bias. Second, it enables us to exploit a more flexible dual stepsize in the convergence analysis. Under some mild conditions, our unified framework preserves $\mathcal{O}(1/T)$ sublinear convergence. Additionally, we establish the linear convergence under error bound conditions. Finally, numerical experiments demonstrate the efficacy of the new algorithm for some nonsmooth and nonconvex problems.

The evolution of wireless communications has been significantly influenced by remarkable advancements in multiple access (MA) technologies over the past five decades, shaping the landscape of modern connectivity. Within this context, a comprehensive tutorial review is presented, focusing on representative MA techniques developed over the past 50 years. The following areas are explored: i) The foundational principles and information-theoretic capacity limits of power-domain non-orthogonal multiple access (NOMA) are characterized, along with its extension to multiple-input multiple-output (MIMO)-NOMA. ii) Several MA transmission schemes exploiting the spatial domain are investigated, encompassing both conventional space-division multiple access (SDMA)/MIMO-NOMA systems and near-field MA systems utilizing spherical-wave propagation models. iii) The application of NOMA to integrated sensing and communications (ISAC) systems is studied. This includes an introduction to typical NOMA-based downlink/uplink ISAC frameworks, followed by an evaluation of their performance limits using a mutual information (MI)-based analytical framework. iv) Major issues and research opportunities associated with the integration of MA with other emerging technologies are identified to facilitate MA in next-generation networks, i.e., next-generation multiple access (NGMA). Throughout the paper, promising directions are highlighted to inspire future research endeavors in the realm of MA and NGMA., Comment: 43 pages, 38 figures; Submitted to Proceedings of the IEEE

In this paper, we present Gauss's law-preserving spectral methods and their efficient solution algorithms for curl-curl source and eigenvalue problems in two and three dimensions arising from Maxwell's equations. Arbitrary order $H(curl)$-conforming spectral basis functions in two and three dimensions are firstly proposed using compact combination of Legendre polynomials. A mixed formulation involving a Lagrange multiplier is then adopted to preserve the Gauss's law in the weak sense. To overcome the bottleneck of computational efficiency caused by the saddle-point nature of the mixed scheme, we present highly efficient solution algorithms based on reordering and decoupling of the resultant linear algebraic system and numerical eigen-decomposition of one dimensional mass matrix. The proposed solution algorithms are direct methods requiring only several matrix-matrix or matrix-tensor products of $N$-by-$N$ matrices, where $N$ is the highest polynomial order in each direction. Compared with other direct methods, the computational complexities are reduced from $O(N^6)$ and $O(N^9)$ to $O(N^3)$ and $O(N^4)$ with small and constant pre-factors for 2D and 3D cases, respectively, and can further be accelerated to $O(N^{2.807})$ and $O(N^{3.807})$, when boosted with the Strassen's matrix multiplication algorithm. Moreover, these algorithms strictly obey the Helmholtz-Hodge decomposition, thus totally eliminate the spurious eigen-modes of non-physical zero eigenvalues. Extensions of the proposed methods and algorithms to problems in complex geometries with variable coefficients and inhomogeneous boundary conditions are discussed to deal with more general situations. Ample numerical examples for solving Maxwell's source and eigenvalue problems are presented to demonstrate the accuracy and efficiency of the proposed methods.

Reconstructing natural language from non-invasive electroencephalography (EEG) holds great promise as a language decoding technology for brain-computer interfaces (BCIs). However, EEG-based language decoding is still in its nascent stages, facing several technical issues such as: 1) Absence of a hybrid strategy that can effectively integrate cross-modality (between EEG and text) self-learning with intra-modality self-reconstruction of EEG features or textual sequences; 2) Under-utilization of large language models (LLMs) to enhance EEG-based language decoding. To address above issues, we propose the Contrastive EEG-Text Masked Autoencoder (CET-MAE), a novel model that orchestrates compound self-supervised learning across and within EEG and text through a dedicated multi-stream encoder. Furthermore, we develop a framework called E2T-PTR (EEG-to-Text decoding using Pretrained Transferable Representations), which leverages pre-trained modules alongside the EEG stream from CET-MAE and further enables an LLM (specifically BART) to decode text from EEG sequences. Comprehensive experiments conducted on the popular text-evoked EEG database, ZuCo, demonstrate the superiority of E2T-PTR, which outperforms the state-of-the-art in ROUGE-1 F1 and BLEU-4 scores by 8.34% and 32.21%, respectively. These results indicate significant advancements in the field and underscores the proposed framework's potential to enable more powerful and widespread BCI applications., Comment: 8 pages (excluding references), accepted by ACL 2024 Main Conference